1
1
openmpi/orte/mca/snapc/full/snapc_full_global.c

2707 строки
97 KiB
C
Исходник Обычный вид История

/*
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* Copyright (c) 2004-2010 The Trustees of Indiana University.
* All rights reserved.
* Copyright (c) 2004-2011 The Trustees of the University of Tennessee.
* All rights reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2007 Evergrid, Inc. All rights reserved.
*
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "orte_config.h"
#include <sys/types.h>
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif /* HAVE_UNISTD_H */
#ifdef HAVE_STRING_H
#include <string.h>
#endif
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
#include "opal/include/opal/prefetch.h"
#include "opal/util/output.h"
#include "opal/util/opal_environ.h"
#include "opal/util/basename.h"
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
#include "opal/util/show_help.h"
#include "orte/mca/mca.h"
#include "opal/mca/base/base.h"
#include "opal/mca/crs/crs.h"
#include "opal/mca/crs/base/base.h"
#include "orte/util/name_fns.h"
#include "orte/util/proc_info.h"
#include "orte/runtime/orte_globals.h"
#include "opal/dss/dss.h"
#include "orte/mca/rml/rml.h"
#include "orte/mca/rml/rml_types.h"
#include "orte/mca/rmaps/rmaps.h"
#include "orte/mca/rmaps/rmaps_types.h"
#include "orte/mca/plm/plm.h"
#include "orte/mca/grpcomm/grpcomm.h"
#include "orte/runtime/orte_wait.h"
#include "orte/mca/errmgr/errmgr.h"
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
#include "orte/mca/errmgr/base/base.h"
#include "orte/mca/snapc/snapc.h"
#include "orte/mca/snapc/base/base.h"
#include "snapc_full.h"
#include MCA_timer_IMPLEMENTATION_HEADER
/************************************
* Locally Global vars & functions :)
************************************/
#define INC_SEQ_NUM() \
{ \
if(orte_snapc_base_store_only_one_seq) { \
orte_snapc_base_snapshot_seq_number = 0; \
} else { \
orte_snapc_base_snapshot_seq_number++; \
} \
}
2009-09-22 22:26:12 +04:00
static orte_jobid_t current_global_jobid = ORTE_JOBID_INVALID;
static orte_snapc_base_global_snapshot_t global_snapshot;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static int current_total_orteds = 0;
static bool updated_job_to_running;
static int current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_NONE;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static bool cleanup_on_establish = false;
static bool global_coord_has_local_children = false;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static bool currently_migrating = false;
static opal_list_t *migrating_procs = NULL;
static int global_init_job_structs(void);
2009-09-22 22:26:12 +04:00
static int global_refresh_job_structs(void);
static bool snapc_orted_recv_issued = false;
2009-09-22 22:26:12 +04:00
static bool is_orte_checkpoint_connected = false;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static bool is_app_checkpointable = false;
static int snapc_full_global_start_listener(void);
static int snapc_full_global_stop_listener(void);
static void snapc_full_global_orted_recv(int status,
orte_process_name_t* sender,
opal_buffer_t* buffer,
orte_rml_tag_t tag,
void* cbdata);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static void snapc_full_process_restart_proc_info_cmd(orte_process_name_t* sender,
opal_buffer_t* buffer);
static void snapc_full_process_request_op_cmd(orte_process_name_t* sender,
2009-09-22 22:26:12 +04:00
opal_buffer_t* buffer);
/*** Command Line Interactions */
static orte_process_name_t orte_checkpoint_sender;
static bool snapc_cmdline_recv_issued = false;
static int snapc_full_global_start_cmdline_listener(void);
static int snapc_full_global_stop_cmdline_listener(void);
static void snapc_full_global_cmdline_recv(int status,
orte_process_name_t* sender,
opal_buffer_t* buffer,
orte_rml_tag_t tag,
void* cbdata);
static int snapc_full_establish_snapshot_dir(bool empty_metadata);
/*** */
2009-09-22 22:26:12 +04:00
static int snapc_full_global_checkpoint(opal_crs_base_ckpt_options_t *options);
static int snapc_full_global_notify_checkpoint(orte_jobid_t jobid,
opal_crs_base_ckpt_options_t *options);
static int orte_snapc_full_global_set_job_ckpt_info( orte_jobid_t jobid,
int ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_sstore_base_handle_t handle,
2009-09-22 22:26:12 +04:00
bool quick,
opal_crs_base_ckpt_options_t *options);
int global_coord_job_state_update(orte_jobid_t jobid,
int job_ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_sstore_base_handle_t handle,
2009-09-22 22:26:12 +04:00
opal_crs_base_ckpt_options_t *options);
static void snapc_full_process_job_update_cmd(orte_process_name_t* sender,
opal_buffer_t* buffer,
bool quick);
static int snapc_full_process_orted_update_cmd(orte_process_name_t* sender,
opal_buffer_t* buffer,
bool quick);
static orte_snapc_full_orted_snapshot_t *find_orted_snapshot(orte_process_name_t *name );
static int snapc_full_global_get_min_state(void);
static int write_out_global_metadata(void);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static int orte_snapc_full_global_reset_coord(void);
/*
* Timer stuff
*/
static void snapc_full_set_time(int idx);
static void snapc_full_display_all_timers(void);
static void snapc_full_display_recovered_timers(void);
static void snapc_full_clear_timers(void);
static double snapc_full_get_time(void);
static void snapc_full_display_indv_timer_core(double diff, char *str);
#define SNAPC_FULL_TIMER_START 0
#define SNAPC_FULL_TIMER_RUNNING 1
#define SNAPC_FULL_TIMER_FIN_LOCAL 2
#define SNAPC_FULL_TIMER_SS_SYNC 3
#define SNAPC_FULL_TIMER_ESTABLISH 4
#define SNAPC_FULL_TIMER_RECOVERED 5
#define SNAPC_FULL_TIMER_MAX 6
static double timer_start[SNAPC_FULL_TIMER_MAX];
#define SNAPC_FULL_CLEAR_TIMERS() \
{ \
if(OPAL_UNLIKELY(orte_snapc_full_timing_enabled)) { \
snapc_full_clear_timers(); \
} \
}
#define SNAPC_FULL_SET_TIMER(idx) \
{ \
if(OPAL_UNLIKELY(orte_snapc_full_timing_enabled)) { \
snapc_full_set_time(idx); \
} \
}
#define SNAPC_FULL_DISPLAY_ALL_TIMERS() \
{ \
if(OPAL_UNLIKELY(orte_snapc_full_timing_enabled)) { \
snapc_full_display_all_timers(); \
} \
}
#define SNAPC_FULL_DISPLAY_RECOVERED_TIMER() \
{ \
if(OPAL_UNLIKELY(orte_snapc_full_timing_enabled)) { \
snapc_full_display_recovered_timers(); \
} \
}
/*
* Progress
*/
static void snapc_full_report_progress(orte_snapc_full_orted_snapshot_t *orted_snapshot,
int total,
int min_state);
static int report_progress_cur_loc_finished = 0;
static double report_progress_last_reported_loc_finished = 0;
#define SNAPC_FULL_REPORT_PROGRESS(orted, total, min_state) \
{ \
if(OPAL_UNLIKELY(orte_snapc_full_progress_meter > 0)) { \
snapc_full_report_progress(orted, total, min_state); \
} \
}
/************************
* Function Definitions
************************/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
int global_coord_init(void)
{
2009-09-22 22:26:12 +04:00
current_global_jobid = ORTE_JOBID_INVALID;
orte_snapc_base_snapshot_seq_number = -1;
orte_checkpoint_sender = orte_name_invalid;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_CLEAR_TIMERS();
2009-09-22 22:26:12 +04:00
return ORTE_SUCCESS;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
int global_coord_finalize(void)
{
2009-09-22 22:26:12 +04:00
current_global_jobid = ORTE_JOBID_INVALID;
orte_snapc_base_snapshot_seq_number = -1;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_CLEAR_TIMERS();
2009-09-22 22:26:12 +04:00
return ORTE_SUCCESS;
}
int global_coord_setup_job(orte_jobid_t jobid) {
int ret, exit_status = ORTE_SUCCESS;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_job_t *jdata = NULL;
/*
* Only allow one job at a time.
*
* It is possible to pass through this function twice since HNP may also be
* a local daemon. So it may be both a global and local coordinator.
* Global: orte_plm_base_setup_job()
* Local : odls_default_module.c
*/
/* Global Coordinator pass */
2009-09-22 22:26:12 +04:00
if( ORTE_JOBID_INVALID == current_global_jobid ) {
current_global_jobid = jobid;
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Setup job %s as the Global Coordinator\n",
ORTE_JOBID_PRINT(jobid)));
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_CLEAR_TIMERS();
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_START);
}
/* Local Coordinator pass - Always happens after global coordinator pass */
else if ( jobid == current_global_jobid ) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/* look up job data object */
if (NULL == (jdata = orte_get_job_data_object(current_global_jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_FLAG_TEST(jdata, ORTE_JOB_FLAG_RESTART)) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Restarting Job %s...",
ORTE_JOBID_PRINT(jobid)));
SNAPC_FULL_CLEAR_TIMERS();
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_START);
if( ORTE_SUCCESS != (ret = global_refresh_job_structs()) ) {
ORTE_ERROR_LOG(ret);
return ret;
}
if( ORTE_SNAPC_LOCAL_COORD_TYPE == (orte_snapc_coord_type & ORTE_SNAPC_LOCAL_COORD_TYPE) ) {
return local_coord_setup_job(jobid);
}
return ORTE_SUCCESS;
}
/* If there are no local children, do not become a local coordinator */
if( !global_coord_has_local_children ) {
return ORTE_SUCCESS;
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Setup job %s as the Local Coordinator\n",
ORTE_JOBID_PRINT(jobid)));
orte_snapc_coord_type |= ORTE_SNAPC_LOCAL_COORD_TYPE;
return local_coord_setup_job(jobid);
}
/* Only allow one job at a time */
else {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) Setup of job %s Failed! Already setup job %s\n",
ORTE_JOBID_PRINT(jobid), ORTE_JOBID_PRINT(current_global_jobid));
ORTE_ERROR_LOG(ORTE_ERROR);
return ORTE_ERROR;
}
/*
* Start out with a sequence number just below the first
* This will be incremented when we checkpoint
*/
orte_snapc_base_snapshot_seq_number = -1;
/*
* Allocate structure to track node status
*/
if( ORTE_SUCCESS != (ret = global_init_job_structs()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/*
* Setup Global Coordinator command processing listener
*/
if( ORTE_SUCCESS != (ret = snapc_full_global_start_listener()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/*
* Setup command line tool checkpoint request listener
*/
if( ORTE_SUCCESS != (ret = snapc_full_global_start_cmdline_listener()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/*
* If requested pre-establish the global snapshot directory
*/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
#if 0
if(orte_snapc_base_establish_global_snapshot_dir) {
opal_output(0, "Global) Error: Pre-establishment of snapshot directory currently not supported!");
ORTE_ERROR_LOG(ORTE_ERR_NOT_SUPPORTED);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Pre-establish the global snapshot directory\n"));
if( ORTE_SUCCESS != (ret = snapc_full_establish_snapshot_dir(true))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
#endif
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Finished setup of job %s ",
ORTE_JOBID_PRINT(jobid)));
cleanup:
return exit_status;
}
int global_coord_release_job(orte_jobid_t jobid) {
int ret, exit_status = ORTE_SUCCESS;
/*
* Make sure we are not waiting on a checkpoint to complete
*/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( is_orte_checkpoint_connected ) {
if( ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
ORTE_SNAPC_CKPT_STATE_ERROR)) ) {
ORTE_ERROR_LOG(ret);
}
}
/*
* Clean up listeners
*/
if( ORTE_SUCCESS != (ret = snapc_full_global_stop_cmdline_listener()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
}
if( ORTE_SUCCESS != (ret = snapc_full_global_stop_listener()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
}
OBJ_DESTRUCT(&global_snapshot);
return exit_status;
}
2009-09-22 22:26:12 +04:00
int global_coord_start_ckpt(orte_snapc_base_quiesce_t *datum)
{
int ret, exit_status = ORTE_SUCCESS;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_std_cntr_t i_proc;
orte_proc_t *proc = NULL;
orte_proc_t *new_proc = NULL;
opal_list_item_t *item = NULL;
2009-09-22 22:26:12 +04:00
opal_crs_base_ckpt_options_t *options = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
char *tmp_str = NULL;
2009-09-22 22:26:12 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Starting checkpoint (internally requested)"));
orte_checkpoint_sender = orte_name_invalid;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* If migrating
*/
if( datum->migrating ) {
currently_migrating = true;
if( NULL != migrating_procs ) {
while( NULL != (item = opal_list_remove_first(migrating_procs)) ) {
proc = (orte_proc_t*)item;
OBJ_RELEASE(proc);
}
} else {
migrating_procs = OBJ_NEW(opal_list_t);
}
/*
* Copy over the procs into a list
*/
for(i_proc = 0; i_proc < opal_pointer_array_get_size(&(datum->migrating_procs)); ++i_proc) {
proc = (orte_proc_t*)opal_pointer_array_get_item(&(datum->migrating_procs), i_proc);
if( NULL == proc ) {
continue;
}
new_proc = OBJ_NEW(orte_proc_t);
new_proc->name.jobid = proc->name.jobid;
new_proc->name.vpid = proc->name.vpid;
new_proc->node = OBJ_NEW(orte_node_t);
new_proc->node->name = proc->node->name;
opal_list_append(migrating_procs, &new_proc->super);
OBJ_RETAIN(new_proc);
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) SnapC Migrating Processes: (%d procs) [Updated]\n",
(int)opal_list_get_size(migrating_procs) ));
for (item = opal_list_get_first(migrating_procs);
item != opal_list_get_end(migrating_procs);
item = opal_list_get_next(item)) {
new_proc = (orte_proc_t*)item;
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"\t\"%s\" [%s]\n",
ORTE_NAME_PRINT(&new_proc->name),new_proc->node->name));
}
}
2009-09-22 22:26:12 +04:00
/*************************
* Kick off the checkpoint (local coord will release the processes)
*************************/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
options = OBJ_NEW(opal_crs_base_ckpt_options_t);
2009-09-22 22:26:12 +04:00
if( ORTE_SUCCESS != (ret = snapc_full_global_checkpoint(options) ) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/*
* Wait for checkpoint to locally finish on all nodes
*/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
while(((currently_migrating && current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_MIGRATING) ||
(!currently_migrating && current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL)) &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ESTABLISHED &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_RECOVERED &&
2009-09-22 22:26:12 +04:00
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_NONE ) {
opal_progress();
}
/*
* Update the quiesce structure with the handle
*/
datum->snapshot = OBJ_NEW(orte_snapc_base_global_snapshot_t);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
datum->ss_handle = global_snapshot.ss_handle;
datum->ss_snapshot = OBJ_NEW(orte_sstore_base_global_snapshot_info_t);
if( ORTE_SUCCESS != (ret = orte_sstore.request_global_snapshot_data(&(datum->ss_handle), datum->ss_snapshot)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/* JJH Is the snapc structure useful with the sstore structure ??? */
orte_sstore.get_attr(global_snapshot.ss_handle,
SSTORE_METADATA_GLOBAL_SNAP_SEQ,
&tmp_str);
datum->epoch = atoi(tmp_str);
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( NULL != tmp_str ) {
free(tmp_str);
tmp_str = NULL;
2009-09-22 22:26:12 +04:00
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
2009-09-22 22:26:12 +04:00
cleanup:
if( NULL != options ) {
OBJ_RELEASE(options);
options = NULL;
}
return exit_status;
}
int global_coord_end_ckpt(orte_snapc_base_quiesce_t *datum)
{
int ret, exit_status = ORTE_SUCCESS;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
opal_list_item_t* item = NULL;
2009-09-22 22:26:12 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) Finishing checkpoint (internally requested) [%3d]",
current_job_ckpt_state));
if( currently_migrating ) {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) End Ckpt: Flush the modex cached data\n"));
/* TODO: You can't pass NULL as the identifier - what you'll need to do is
* close all open dstore handles, and then open the ones you need
*/
#if 0
if (OPAL_SUCCESS != (ret = opal_dstore.remove(NULL, NULL))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
#endif
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_ESTABLISH);
if( ORTE_SUCCESS != (ret = orte_snapc_full_global_set_job_ckpt_info(current_global_jobid,
ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL,
global_snapshot.ss_handle,
true, NULL) ) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
while(current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_RECOVERED &&
2009-09-22 22:26:12 +04:00
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_NONE ) {
opal_progress();
}
/*
* Update the job structure since processes may have moved around
*/
if( ORTE_SUCCESS != (ret = global_refresh_job_structs()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Finished checkpoint (internally requested) [%d]",
current_job_ckpt_state));
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( currently_migrating ) {
current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_NONE;
cleanup_on_establish = false;
report_progress_cur_loc_finished = 0;
report_progress_last_reported_loc_finished = 0;
}
2009-09-22 22:26:12 +04:00
cleanup:
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
currently_migrating = false;
if( NULL != migrating_procs ) {
while( NULL != (item = opal_list_remove_first(migrating_procs)) ) {
OBJ_RELEASE(item);
}
OBJ_RELEASE(migrating_procs);
migrating_procs = NULL;
}
2009-09-22 22:26:12 +04:00
return exit_status;
}
/******************
* Local functions
******************/
static int global_init_job_structs(void)
{
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
orte_snapc_base_local_snapshot_t *app_snapshot = NULL;
opal_list_item_t* orted_item = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_node_t *cur_node = NULL;
orte_job_map_t *map = NULL;
orte_job_t *jdata = NULL;
orte_proc_t **procs = NULL;
orte_std_cntr_t i = 0;
orte_vpid_t p = 0;
orte_ns_cmp_bitmask_t mask;
bool found = false;
/* look up job data object */
if (NULL == (jdata = orte_get_job_data_object(current_global_jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
OBJ_CONSTRUCT(&global_snapshot, orte_snapc_base_global_snapshot_t);
map = jdata->map;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for (i=0; i < map->nodes->size; i++) {
if (NULL == (cur_node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
procs = (orte_proc_t**)cur_node->procs->addr;
/*
* Look out for duplicates
* JJH: Should not happen, but does if rmaps get a bug in setting up the map.
*/
found = false;
for(orted_item = opal_list_get_first(&(global_snapshot.local_snapshots));
orted_item != opal_list_get_end(&(global_snapshot.local_snapshots));
orted_item = opal_list_get_next(orted_item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)orted_item;
/*
* Is in list?
*/
if(OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&(cur_node->daemon->name),
&(orted_snapshot->process_name) )) {
found = true;
break;
}
}
if( found ) {
OPAL_OUTPUT_VERBOSE((1, mca_snapc_full_component.super.output_handle,
"Global) [%d] Found Daemon %s with %d procs - Duplicate!! - Should not happen!",
i, ORTE_NAME_PRINT(&(cur_node->daemon->name)), cur_node->num_procs));
continue;
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) [%d] Found Daemon %s with %d procs",
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
i, ORTE_NAME_PRINT(&(cur_node->daemon->name)), cur_node->num_procs));
orted_snapshot = OBJ_NEW(orte_snapc_full_orted_snapshot_t);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orted_snapshot->process_name.jobid = cur_node->daemon->name.jobid;
orted_snapshot->process_name.vpid = cur_node->daemon->name.vpid;
mask = ORTE_NS_CMP_JOBID;
if (OPAL_EQUAL ==
orte_util_compare_name_fields(mask, &orted_snapshot->process_name, ORTE_PROC_MY_NAME)) {
global_coord_has_local_children = true;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for(p = 0; p < cur_node->num_procs; ++p) {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) \t [%d] Found Process %s on Daemon %s",
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
p, ORTE_NAME_PRINT(&(procs[p]->name)), ORTE_NAME_PRINT(&(cur_node->daemon->name)) ));
app_snapshot = OBJ_NEW(orte_snapc_base_local_snapshot_t);
app_snapshot->process_name.jobid = procs[p]->name.jobid;
app_snapshot->process_name.vpid = procs[p]->name.vpid;
opal_list_append(&(orted_snapshot->super.local_snapshots), &(app_snapshot->super));
}
opal_list_append(&global_snapshot.local_snapshots, &(orted_snapshot->super.super));
}
return ORTE_SUCCESS;
}
2009-09-22 22:26:12 +04:00
static int global_refresh_job_structs(void)
{
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
orte_snapc_base_local_snapshot_t *app_snapshot = NULL;
opal_list_item_t* orted_item = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
opal_list_item_t* app_item = NULL;
opal_list_item_t* item = NULL;
orte_node_t *cur_node = NULL;
2009-09-22 22:26:12 +04:00
orte_job_map_t *map = NULL;
orte_job_t *jdata = NULL;
orte_proc_t **procs = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_proc_t *new_proc = NULL;
2009-09-22 22:26:12 +04:00
orte_std_cntr_t i = 0;
orte_vpid_t p = 0;
bool found = false;
orte_ns_cmp_bitmask_t mask;
2009-09-22 22:26:12 +04:00
/* look up job data object */
if (NULL == (jdata = orte_get_job_data_object(current_global_jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Refreshing Job Structures... [%3d]",
current_job_ckpt_state));
if( NULL != migrating_procs ) {
for (item = opal_list_get_first(migrating_procs);
item != opal_list_get_end(migrating_procs);
item = opal_list_get_next(item)) {
new_proc = (orte_proc_t*)item;
/*
* Look through all daemons
*/
found = false;
for(orted_item = opal_list_get_first(&(global_snapshot.local_snapshots));
orted_item != opal_list_get_end(&(global_snapshot.local_snapshots));
orted_item = opal_list_get_next(orted_item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)orted_item;
/*
* Look through all processes tracked by this daemon
*/
for(app_item = opal_list_get_first(&(orted_snapshot->super.local_snapshots));
app_item != opal_list_get_end(&(orted_snapshot->super.local_snapshots));
app_item = opal_list_get_next(app_item) ) {
app_snapshot = (orte_snapc_base_local_snapshot_t*)app_item;
if(OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&(new_proc->name),
&(app_snapshot->process_name) )) {
found = true;
opal_list_remove_item(&(orted_snapshot->super.local_snapshots), app_item);
break;
}
}
if( found ) {
break;
}
}
}
}
/*
* First make sure that all of the orted's have the proper number of
* children, if no children, then stop tracking.
*/
2009-09-22 22:26:12 +04:00
map = jdata->map;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for(orted_item = opal_list_get_first(&(global_snapshot.local_snapshots));
orted_item != opal_list_get_end(&(global_snapshot.local_snapshots));
orted_item = opal_list_get_next(orted_item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)orted_item;
/* Make sure this orted is in the map */
found = false;
for (i=0; i < map->nodes->size; i++) {
if (NULL == (cur_node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
if(OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&(cur_node->daemon->name),
&(orted_snapshot->process_name) )) {
found = true;
break;
}
}
/* If not, then remove all processes, keep ref. we might reuse it later */
if( !found ) {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Found Empty Daemon %s not in map (Refresh)",
ORTE_NAME_PRINT(&(orted_snapshot->process_name)) ));
while( NULL != (item = opal_list_remove_first(&(orted_snapshot->super.local_snapshots))) ) {
OBJ_RELEASE(item);
}
}
}
2009-09-22 22:26:12 +04:00
/*
* Look for new nodes
*/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for (i=0; i < map->nodes->size; i++) {
if (NULL == (cur_node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
procs = (orte_proc_t**)cur_node->procs->addr;
2009-09-22 22:26:12 +04:00
/*
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* See if we are already tracking it, if so refresh it
* (This daemon could have been restarted, and processes migrated back to it)
2009-09-22 22:26:12 +04:00
*/
found = false;
for(orted_item = opal_list_get_first(&(global_snapshot.local_snapshots));
orted_item != opal_list_get_end(&(global_snapshot.local_snapshots));
orted_item = opal_list_get_next(orted_item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)orted_item;
if(OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
&(cur_node->daemon->name),
2009-09-22 22:26:12 +04:00
&(orted_snapshot->process_name) )) {
found = true;
break;
}
}
if( found ) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) [%d] Found Daemon %s with %d procs (Refresh)",
i, ORTE_NAME_PRINT(&(cur_node->daemon->name)), cur_node->num_procs));
/* Remove all old processes */
while( NULL != (item = opal_list_remove_first(&(orted_snapshot->super.local_snapshots))) ) {
OBJ_RELEASE(item);
}
/* Add back new processes (a bit of overkill, sure, but it works) */
for(p = 0; p < cur_node->num_procs; ++p) {
if( NULL == procs[p] ) {
continue;
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) \t [%d] Found Process %s on Daemon %s",
p, ORTE_NAME_PRINT(&(procs[p]->name)), ORTE_NAME_PRINT(&(cur_node->daemon->name)) ));
app_snapshot = OBJ_NEW(orte_snapc_base_local_snapshot_t);
app_snapshot->process_name.jobid = procs[p]->name.jobid;
app_snapshot->process_name.vpid = procs[p]->name.vpid;
opal_list_append(&(orted_snapshot->super.local_snapshots), &(app_snapshot->super));
}
2009-09-22 22:26:12 +04:00
continue;
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) [%d] Found Daemon %s with %d procs",
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
i, ORTE_NAME_PRINT(&(cur_node->daemon->name)), cur_node->num_procs));
2009-09-22 22:26:12 +04:00
orted_snapshot = OBJ_NEW(orte_snapc_full_orted_snapshot_t);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orted_snapshot->process_name.jobid = cur_node->daemon->name.jobid;
orted_snapshot->process_name.vpid = cur_node->daemon->name.vpid;
mask = ORTE_NS_CMP_ALL;
2009-09-22 22:26:12 +04:00
if (OPAL_EQUAL ==
orte_util_compare_name_fields(mask, &orted_snapshot->process_name, ORTE_PROC_MY_NAME)) {
2009-09-22 22:26:12 +04:00
global_coord_has_local_children = true;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for(p = 0; p < cur_node->num_procs; ++p) {
if( NULL == procs[p] ) {
continue;
}
2009-09-22 22:26:12 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) \t [%d] Found Process %s on Daemon %s",
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
p, ORTE_NAME_PRINT(&(procs[p]->name)), ORTE_NAME_PRINT(&(cur_node->daemon->name)) ));
2009-09-22 22:26:12 +04:00
app_snapshot = OBJ_NEW(orte_snapc_base_local_snapshot_t);
app_snapshot->process_name.jobid = procs[p]->name.jobid;
app_snapshot->process_name.vpid = procs[p]->name.vpid;
opal_list_append(&(orted_snapshot->super.local_snapshots), &(app_snapshot->super));
}
opal_list_append(&global_snapshot.local_snapshots, &(orted_snapshot->super.super));
}
return ORTE_SUCCESS;
}
/*****************
* Setup listeners
*****************/
static int snapc_full_global_start_listener(void)
{
if (snapc_orted_recv_issued && ORTE_PROC_IS_HNP) {
return ORTE_SUCCESS;
}
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) Startup Coordinator Channel"));
/*
* Coordinator command listener
*/
orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_SNAPC_FULL,
ORTE_RML_PERSISTENT, snapc_full_global_orted_recv, NULL);
snapc_orted_recv_issued = true;
return ORTE_SUCCESS;
}
static int snapc_full_global_stop_listener(void)
{
if (!snapc_orted_recv_issued && ORTE_PROC_IS_HNP) {
return ORTE_SUCCESS;
}
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) Shutdown Coordinator Channel"));
orte_rml.recv_cancel(ORTE_NAME_WILDCARD, ORTE_RML_TAG_SNAPC_FULL);
snapc_orted_recv_issued = false;
return ORTE_SUCCESS;
}
static int snapc_full_global_start_cmdline_listener(void)
{
if (snapc_cmdline_recv_issued && ORTE_PROC_IS_HNP) {
return ORTE_SUCCESS;
}
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) Startup Command Line Channel"));
/*
* Coordinator command listener
*/
orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_CKPT, 0,
snapc_full_global_cmdline_recv, NULL);
snapc_cmdline_recv_issued = true;
return ORTE_SUCCESS;
}
static int snapc_full_global_stop_cmdline_listener(void)
{
if (!snapc_cmdline_recv_issued && ORTE_PROC_IS_HNP) {
return ORTE_SUCCESS;
}
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) Shutdown Command Line Channel"));
orte_rml.recv_cancel(ORTE_NAME_WILDCARD, ORTE_RML_TAG_CKPT);
snapc_cmdline_recv_issued = false;
return ORTE_SUCCESS;
}
/*****************
* Listener Callbacks
*****************/
static void snapc_full_global_cmdline_recv(int status,
orte_process_name_t* sender,
opal_buffer_t* buffer,
orte_rml_tag_t tag,
void* cbdata)
{
int ret;
orte_snapc_cmd_flag_t command;
orte_std_cntr_t count = 1;
orte_jobid_t jobid;
opal_crs_base_ckpt_options_t *options = NULL;
if( ORTE_RML_TAG_CKPT != tag ) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) Error: Unknown tag: Received a command message from %s (tag = %d).",
ORTE_NAME_PRINT(sender), tag);
ORTE_ERROR_LOG(ORTE_ERR_BAD_PARAM);
return;
}
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command Line: Start a checkpoint operation [Sender = %s]",
ORTE_NAME_PRINT(sender)));
snapc_cmdline_recv_issued = false; /* Not a persistent RML message */
2009-09-22 22:26:12 +04:00
options = OBJ_NEW(opal_crs_base_ckpt_options_t);
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &command, &count, ORTE_SNAPC_CMD))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
/*
* orte_checkpoint has requested that a checkpoint be taken
*/
if (ORTE_SNAPC_GLOBAL_INIT_CMD == command) {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command line requested a checkpoint [command %d]\n",
command));
/*
* Unpack the buffer from the orte_checkpoint command
*/
2009-09-22 22:26:12 +04:00
if( ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_init_cmd(sender,
buffer,
2009-09-22 22:26:12 +04:00
options,
&jobid)) ) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_checkpoint_sender = *sender;
is_orte_checkpoint_connected = true;
/*
* If the application is not ready for a checkpoint,
* then send back an error.
*/
if( !is_app_checkpointable ) {
OPAL_OUTPUT_VERBOSE((1, mca_snapc_full_component.super.output_handle,
"Global) request_cmd(): Checkpointing currently disabled, rejecting request"));
if( ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
0,
ORTE_SNAPC_CKPT_STATE_ERROR))) {
ORTE_ERROR_LOG(ret);
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_checkpoint_sender = orte_name_invalid;
is_orte_checkpoint_connected = false;
/* Reset the listener */
if( ORTE_SUCCESS != (ret = snapc_full_global_start_cmdline_listener() ) ){
ORTE_ERROR_LOG(ret);
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
goto cleanup;
}
2009-09-22 22:26:12 +04:00
/*
* If the jobid was specified, and does not match the current job, then fail
*/
if( ORTE_JOBID_INVALID != jobid && jobid != current_global_jobid) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) Error: Jobid %s does not match the current jobid %s",
ORTE_JOBID_PRINT(jobid), ORTE_JOBID_PRINT(current_global_jobid));
ORTE_ERROR_LOG(ORTE_ERROR);
goto cleanup;
}
/*************************
* Kick off the checkpoint
*************************/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_CLEAR_TIMERS();
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_START);
2009-09-22 22:26:12 +04:00
if( ORTE_SUCCESS != (ret = snapc_full_global_checkpoint(options) ) ) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
}
/*
* Terminate the connection (Not currently implemented)
*/
else if (ORTE_SNAPC_GLOBAL_TERM_CMD == command) {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command line requested to terminate connection (command %d)\n",
command));
ORTE_ERROR_LOG(ORTE_ERR_NOT_SUPPORTED);
goto cleanup;
}
/*
* Unknown command
*/
else {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command line sent an unknown command (command %d)\n",
command));
ORTE_ERROR_LOG(ORTE_ERR_NOT_SUPPORTED);
goto cleanup;
}
cleanup:
2009-09-22 22:26:12 +04:00
if( NULL != options ) {
OBJ_RELEASE(options);
options = NULL;
}
return;
}
void snapc_full_global_orted_recv(int status,
orte_process_name_t* sender,
opal_buffer_t* buffer,
orte_rml_tag_t tag,
void* cbdata)
{
int ret;
orte_snapc_full_cmd_flag_t command;
orte_std_cntr_t count;
static int num_inside = 0;
if( ORTE_RML_TAG_SNAPC_FULL != tag ) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) Error: Unknown tag: Received a command message from %s (tag = %d).",
ORTE_NAME_PRINT(sender), tag);
ORTE_ERROR_LOG(ORTE_ERR_BAD_PARAM);
return;
}
/*
* This is a message from a Local Coordinator
*/
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) Receive a command message from %s.",
ORTE_NAME_PRINT(sender)));
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &command, &count, ORTE_SNAPC_FULL_CMD))) {
ORTE_ERROR_LOG(ret);
return;
}
++num_inside;
switch (command) {
case ORTE_SNAPC_FULL_UPDATE_JOB_STATE_QUICK_CMD:
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command: Job State Update (quick)"));
snapc_full_process_job_update_cmd(sender, buffer, true);
break;
case ORTE_SNAPC_FULL_UPDATE_JOB_STATE_CMD:
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command: Job State Update"));
snapc_full_process_job_update_cmd(sender, buffer, false);
break;
case ORTE_SNAPC_FULL_UPDATE_ORTED_STATE_QUICK_CMD:
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command: Daemon State Update (quick)"));
snapc_full_process_orted_update_cmd(sender, buffer, true);
break;
case ORTE_SNAPC_FULL_UPDATE_ORTED_STATE_CMD:
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Command: Daemon State Update"));
snapc_full_process_orted_update_cmd(sender, buffer, false);
break;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
case ORTE_SNAPC_FULL_RESTART_PROC_INFO:
2009-09-22 22:26:12 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) Command: Update hostname/pid associations"));
snapc_full_process_restart_proc_info_cmd(sender, buffer);
2009-09-22 22:26:12 +04:00
break;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
case ORTE_SNAPC_FULL_REQUEST_OP_CMD:
2009-09-22 22:26:12 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) Command: Request Op"));
2009-09-22 22:26:12 +04:00
snapc_full_process_request_op_cmd(sender, buffer);
2009-09-22 22:26:12 +04:00
break;
default:
ORTE_ERROR_LOG(ORTE_ERR_VALUE_OUT_OF_BOUNDS);
}
return;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static void snapc_full_process_request_op_cmd(orte_process_name_t* sender,
2009-09-22 22:26:12 +04:00
opal_buffer_t* sbuffer)
{
int ret;
orte_std_cntr_t count = 1;
orte_jobid_t jobid;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
int op_event, op_state;
2009-09-22 22:26:12 +04:00
opal_crs_base_ckpt_options_t *options = NULL;
opal_buffer_t *buffer = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_snapc_full_cmd_flag_t command = ORTE_SNAPC_FULL_REQUEST_OP_CMD;
int seq_num = -1, i;
char * global_handle = NULL, *tmp_str = NULL;
orte_snapc_base_request_op_t *datum = NULL;
2009-09-22 22:26:12 +04:00
orte_checkpoint_sender = orte_name_invalid;
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &jobid, &count, ORTE_JOBID))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &op_event, &count, OPAL_INT))) {
2009-09-22 22:26:12 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Op Code %2d\n",
op_event));
/************************************
* Application have been initialized, and are ready for checkpointing
************************************/
if( ORTE_SNAPC_OP_INIT == op_event ) {
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Checkpointing Enabled (%2d)\n",
op_event));
is_app_checkpointable = true;
}
/************************************
* Application is finalizing, and no longer ready for checkpointing.
************************************/
else if( ORTE_SNAPC_OP_FIN == op_event ) {
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Checkpointing Disabled (%2d)\n",
op_event));
is_app_checkpointable = false;
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Wait for any ongoing checkpoints to finish
*/
if( current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_NONE ) {
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Wait for ongoing checkpoint to complete..."));
while( current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_NONE ) {
opal_progress();
}
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Tell application that it is now ok to finailze
*/
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Send Finalize ACK to the job"));
2009-09-22 22:26:12 +04:00
buffer = OBJ_NEW(opal_buffer_t);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
op_event = ORTE_SNAPC_OP_FIN_ACK;
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &op_event, 1, OPAL_INT))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if (ORTE_SUCCESS != (ret = orte_rml.send_buffer_nb(sender, buffer, ORTE_RML_TAG_SNAPC_FULL,
orte_rml_send_callback, NULL))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
/* buffer should not be released here; the callback releases it */
buffer = NULL;
2009-09-22 22:26:12 +04:00
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/************************************
* Start a checkpoint operation
************************************/
else if( ORTE_SNAPC_OP_CHECKPOINT == op_event ) {
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Starting checkpoint (%2d)\n",
op_event));
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
options = OBJ_NEW(opal_crs_base_ckpt_options_t);
if( ORTE_SUCCESS != (ret = snapc_full_global_checkpoint(options) ) ) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Wait for the operation to complete
*/
while( current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_NONE ) {
opal_progress();
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( ORTE_SNAPC_CKPT_STATE_ERROR == current_job_ckpt_state ) {
op_state = -1;
} else {
op_state = 0;
}
/*
* Tell the sender that the operation is finished
*/
buffer = OBJ_NEW(opal_buffer_t);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &op_event, 1, OPAL_INT))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &op_state, 1, OPAL_INT))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if (ORTE_SUCCESS != (ret = orte_rml.send_buffer_nb(sender, buffer, ORTE_RML_TAG_SNAPC_FULL,
orte_rml_send_callback, NULL))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
/* buffer should not be released here; the callback releases it */
buffer = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
}
/************************************
* Start the Restart operation
************************************/
else if( ORTE_SNAPC_OP_RESTART == op_event ) {
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Starting restart (%2d)\n",
op_event));
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &seq_num, &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
orte_snapc_ckpt_state_notify(ORTE_SNAPC_CKPT_STATE_NO_RESTART);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
goto cleanup;
}
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &global_handle, &count, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
orte_snapc_ckpt_state_notify(ORTE_SNAPC_CKPT_STATE_NO_RESTART);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
goto cleanup;
}
/*
* Kick off the restart
*/
if( ORTE_SUCCESS != (ret = orte_errmgr_base_restart_job(current_global_jobid, global_handle, seq_num) ) ) {
ORTE_ERROR_LOG(ret);
orte_snapc_ckpt_state_notify(ORTE_SNAPC_CKPT_STATE_NO_RESTART);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
goto cleanup;
}
}
/************************************
* Start the Migration operation
************************************/
else if( ORTE_SNAPC_OP_MIGRATE == op_event ) {
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Starting migration (%2d)\n",
op_event));
datum = OBJ_NEW(orte_snapc_base_request_op_t);
/*
* Unpack migration information
*/
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &(datum->mig_num), &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
datum->mig_vpids = malloc(sizeof(int) * datum->mig_num);
datum->mig_host_pref = malloc(sizeof(char) * datum->mig_num * OPAL_MAX_PROCESSOR_NAME);
datum->mig_vpid_pref = malloc(sizeof(int) * datum->mig_num);
datum->mig_off_node = malloc(sizeof(int) * datum->mig_num);
for( i = 0; i < datum->mig_num; ++i ) {
(datum->mig_vpids)[i] = 0;
(datum->mig_host_pref)[i][0] = '\0';
(datum->mig_vpid_pref)[i] = 0;
(datum->mig_off_node)[i] = (int)false;
}
for( i = 0; i < datum->mig_num; ++i ) {
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &((datum->mig_vpids)[i]), &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if(NULL != tmp_str ) {
free(tmp_str);
tmp_str = NULL;
}
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &tmp_str, &count, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
strncpy( ((datum->mig_host_pref)[i]), tmp_str, OPAL_MAX_PROCESSOR_NAME);
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &((datum->mig_vpid_pref)[i]), &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(sbuffer, &((datum->mig_off_node)[i]), &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) Migration %3d/%3d: Received Rank %3d - Requested <%s> (%3d) %c\n",
datum->mig_num, i,
(datum->mig_vpids)[i],
(datum->mig_host_pref)[i],
(datum->mig_vpid_pref)[i],
(OPAL_INT_TO_BOOL((datum->mig_off_node)[i]) ? 'T' : 'F')
));
}
/*
* Kick off the migration
*/
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) ------ Kick Off Migration -----"));
if( ORTE_SUCCESS != (ret = orte_errmgr_base_migrate_job(current_global_jobid, datum) ) ) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
/*
* Tell the sender that the operation is finished
*/
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) ------ Finished Migration. Release processes (%15s )-----",
ORTE_NAME_PRINT(sender) ));
buffer = OBJ_NEW(opal_buffer_t);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &op_event, 1, OPAL_INT))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
op_state = 0;
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &op_state, 1, OPAL_INT))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
if (ORTE_SUCCESS != (ret = orte_rml.send_buffer_nb(sender, buffer, ORTE_RML_TAG_SNAPC_FULL,
orte_rml_send_callback, NULL))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
goto cleanup;
}
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) ------ Finished Migration. Released processes (%15s )-----",
ORTE_NAME_PRINT(sender) ));
}
/************************************
* Start the Quiesce operation
************************************/
else if( ORTE_SNAPC_OP_QUIESCE_START == op_event) {
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Starting quiesce (%2d)\n",
op_event));
options = OBJ_NEW(opal_crs_base_ckpt_options_t);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
options->inc_prep_only = true;
if( ORTE_SUCCESS != (ret = snapc_full_global_checkpoint(options) ) ) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
/*
* Wait for quiescence
*/
while( current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_INC_PREPED ) {
opal_progress();
}
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Quiesce_start finished(%2d)\n",
op_event));
}
/************************************
* End the Quiesce operation
************************************/
else if( ORTE_SNAPC_OP_QUIESCE_END == op_event) {
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Ending quiesce (%2d)\n",
op_event));
/*
* Wait for the checkpoint operation to finish
*/
while( current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_ERROR &&
current_job_ckpt_state != ORTE_SNAPC_CKPT_STATE_NONE ) {
opal_progress();
}
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) process_request_op(): Quiesce_end finished(%2d)\n",
op_event));
}
cleanup:
if (NULL != buffer) {
OBJ_RELEASE(buffer);
buffer = NULL;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( NULL != options ) {
OBJ_RELEASE(options);
options = NULL;
}
if(NULL != tmp_str ) {
free(tmp_str);
tmp_str = NULL;
}
return;
}
static int snapc_full_process_orted_update_cmd(orte_process_name_t* sender,
opal_buffer_t* buffer,
bool quick)
{
int ret, exit_status = ORTE_SUCCESS;
orte_std_cntr_t count;
int remote_ckpt_state;
opal_list_item_t* item = NULL;
opal_list_item_t* aitem = NULL;
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
orte_snapc_base_local_snapshot_t *app_snapshot = NULL;
int loc_min_state;
char *state_str = NULL;
orted_snapshot = find_orted_snapshot(sender);
if( NULL == orted_snapshot ) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) Error: Unknown Daemon %s",
2009-09-22 22:26:12 +04:00
ORTE_NAME_PRINT(sender) );
exit_status = ORTE_ERROR;
ORTE_ERROR_LOG(ORTE_ERROR);
goto cleanup;
}
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) Daemon %s: Changed state to:\n",
ORTE_NAME_PRINT(&(orted_snapshot->process_name)) ));
/*
* Unpack the data (quick)
* - state
* Unpack the data (long)
* - state
* - # procs
* - Foreach proc
* - process name
*/
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &remote_ckpt_state, &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
orted_snapshot->state = remote_ckpt_state;
orte_snapc_ckpt_state_str(&state_str, orted_snapshot->state);
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) State: %d (%s)\n",
(int)(orted_snapshot->state), state_str));
free(state_str);
state_str = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/* JJH: Though there is currently no additional information sent in a long
* message versus a small message, keep this logic so that in the
* future it can be easily reused without substantially modifying
* the component.
*/
if( quick ) {
exit_status = ORTE_SUCCESS;
goto post_process;
}
post_process:
loc_min_state = snapc_full_global_get_min_state();
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_REPORT_PROGRESS(orted_snapshot, current_total_orteds, loc_min_state);
/*
* Notify the orte-checkpoint command once we have everyone running.
* No need to broadcast this to everyone since they already know.
*/
if( ORTE_SNAPC_CKPT_STATE_RUNNING == loc_min_state &&
ORTE_SNAPC_CKPT_STATE_RUNNING != current_job_ckpt_state) {
current_job_ckpt_state = loc_min_state;
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_RUNNING);
2009-09-22 22:26:12 +04:00
if( is_orte_checkpoint_connected &&
ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
2009-09-22 22:26:12 +04:00
current_job_ckpt_state)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* If we are just prep'ing the INC, then acknowledge the state change
*/
if( ORTE_SNAPC_CKPT_STATE_INC_PREPED == loc_min_state &&
ORTE_SNAPC_CKPT_STATE_INC_PREPED > current_job_ckpt_state) {
current_job_ckpt_state = loc_min_state;
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) All Processes have finished the INC prep!\n"));
}
2009-09-22 22:26:12 +04:00
/*
* Notify the orte-checkpoint command once we have everyone stopped.
* No need to broadcast this to everyone since they already know.
*/
if( ORTE_SNAPC_CKPT_STATE_STOPPED == loc_min_state &&
ORTE_SNAPC_CKPT_STATE_STOPPED > current_job_ckpt_state) {
current_job_ckpt_state = loc_min_state;
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) All Processes have been stopped!\n"));
if( is_orte_checkpoint_connected &&
ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
current_job_ckpt_state)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
2009-09-22 22:26:12 +04:00
/* orte-checkpoint detaches at this point */
is_orte_checkpoint_connected = false;
/*
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* Synchronize the checkpoint here
2009-09-22 22:26:12 +04:00
*/
write_out_global_metadata();
}
/*
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* If all daemons have finished, let everyone know we are locally finished.
*/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL == loc_min_state &&
ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL > current_job_ckpt_state) {
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_FIN_LOCAL);
if( ORTE_SNAPC_CKPT_STATE_NONE != current_job_ckpt_state ) {
if( loc_min_state == current_job_ckpt_state) {
opal_output(0, "Global) JJH WARNING!!: (%d) == (%d)", loc_min_state, current_job_ckpt_state);
}
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( currently_migrating ) {
write_out_global_metadata();
current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_MIGRATING;
}
else {
current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL;
}
if( NULL != state_str ) {
free(state_str);
}
orte_snapc_ckpt_state_str(&state_str, current_job_ckpt_state);
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) Job State Changed: %d (%s)\n",
(int)current_job_ckpt_state, state_str ));
free(state_str);
state_str = NULL;
if( ORTE_SUCCESS != (ret = orte_snapc_full_global_set_job_ckpt_info(current_global_jobid,
current_job_ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
true, NULL) ) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/*
* Now that we have finished locally,
* - Write out the metadata
* - Sync the snapshot to SStore
* if we are stopping then we have already written out this data.
*/
if( !(global_snapshot.options->stop) && !currently_migrating ) {
write_out_global_metadata();
}
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_ESTABLISH);
if( ORTE_SUCCESS != (ret = orte_snapc_full_global_set_job_ckpt_info(current_global_jobid,
ORTE_SNAPC_CKPT_STATE_ESTABLISHED,
global_snapshot.ss_handle,
true, NULL) ) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
/*
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* If all daemons have confirmed that their local proces are finished
* and we have finished establishing the checkpoint,
* then let the command line tool know and cleanup.
*/
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( ORTE_SNAPC_CKPT_STATE_RECOVERED == loc_min_state &&
ORTE_SNAPC_CKPT_STATE_RECOVERED > current_job_ckpt_state ) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* If this is a job restarting then we do something different
*/
if( current_job_ckpt_state == ORTE_SNAPC_CKPT_STATE_NONE ) {
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
"Global) Job has been successfully restarted"));
/*current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_RECOVERED;*/
orte_snapc_ckpt_state_notify(ORTE_SNAPC_CKPT_STATE_RECOVERED);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for(item = opal_list_get_first(&(global_snapshot.local_snapshots));
item != opal_list_get_end(&(global_snapshot.local_snapshots));
item = opal_list_get_next(item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
orted_snapshot->state = ORTE_SNAPC_CKPT_STATE_NONE;
for(aitem = opal_list_get_first(&(orted_snapshot->super.local_snapshots));
aitem != opal_list_get_end(&(orted_snapshot->super.local_snapshots));
aitem = opal_list_get_next(aitem) ) {
app_snapshot = (orte_snapc_base_local_snapshot_t*)aitem;
app_snapshot->state = ORTE_SNAPC_CKPT_STATE_NONE;
}
}
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_RECOVERED);
SNAPC_FULL_DISPLAY_RECOVERED_TIMER();
orte_snapc_base_has_recovered = true;
is_app_checkpointable = true;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
exit_status = ORTE_SUCCESS;
goto cleanup;
}
/*
* If the checkpoint has not been established yet, then do not clear the
* snapshot structure just yet.
*/
if(ORTE_SNAPC_CKPT_STATE_ESTABLISHED != current_job_ckpt_state ) {
cleanup_on_establish = true;
}
current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_RECOVERED;
if( NULL != state_str ) {
free(state_str);
}
orte_snapc_ckpt_state_str(&state_str, current_job_ckpt_state);
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) Job State Changed: %d (%s)\n",
(int)current_job_ckpt_state, state_str ));
free(state_str);
state_str = NULL;
/*
* Notify the orte-checkpoint command
*/
if( is_orte_checkpoint_connected &&
ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
current_job_ckpt_state)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_RECOVERED);
/*
* If the checkpoint has been established at this point, then cleanup.
*/
if( !cleanup_on_establish && ORTE_SNAPC_CKPT_STATE_RECOVERED == current_job_ckpt_state) {
if( ORTE_SUCCESS != (ret = orte_snapc_full_global_reset_coord()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
}
cleanup:
if( NULL != state_str ) {
free(state_str);
state_str = NULL;
}
return exit_status;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static void snapc_full_process_restart_proc_info_cmd(orte_process_name_t* sender,
opal_buffer_t* buffer)
{
int ret;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_std_cntr_t count;
size_t num_vpids = 0, i;
pid_t tmp_pid;
char * tmp_hostname = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &tmp_hostname, &count, OPAL_STRING))) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) vpid_assoc: Failed to unpack process Hostname from peer %s\n",
ORTE_NAME_PRINT(sender));
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &num_vpids, &count, OPAL_SIZE))) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) vpid_assoc: Failed to unpack num_vpids from peer %s\n",
ORTE_NAME_PRINT(sender));
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
for(i = 0; i < num_vpids; ++i) {
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &tmp_pid, &count, OPAL_PID))) {
opal_output(mca_snapc_full_component.super.output_handle,
"Global) vpid_assoc: Failed to unpack process PID from peer %s\n",
ORTE_NAME_PRINT(sender));
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_coord_restart_proc_info(tmp_pid, tmp_hostname);
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/* stdout may be buffered by the C library so it needs to be flushed so
* that the debugger can read the process info.
*/
fflush(stdout);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
cleanup:
return;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
int global_coord_restart_proc_info(pid_t local_pid, char * local_hostname)
{
printf("MPIR_debug_info) %s:%d\n", local_hostname, local_pid);
return 0;
}
static void snapc_full_process_job_update_cmd(orte_process_name_t* sender,
opal_buffer_t* buffer,
bool quick)
{
int ret;
orte_std_cntr_t count;
orte_jobid_t jobid;
int job_ckpt_state = ORTE_SNAPC_CKPT_STATE_NONE;
2009-09-22 22:26:12 +04:00
opal_crs_base_ckpt_options_t *options = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
bool loc_migrating = false;
size_t loc_num_procs = 0;
orte_proc_t *proc = NULL;
size_t i;
orte_sstore_base_handle_t ss_handle;
/*
2009-09-22 22:26:12 +04:00
* Unpack the data (quick)
* - jobid
* - ckpt_state
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* - sstore_handle
2009-09-22 22:26:12 +04:00
* Unpack the data (long)
* - jobid
* - ckpt_state
2009-09-22 22:26:12 +04:00
* - ckpt_options
*/
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &jobid, &count, ORTE_JOBID))) {
ORTE_ERROR_LOG(ret);
return;
}
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &job_ckpt_state, &count, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
return;
}
if( !quick ) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if (ORTE_SUCCESS != (ret = orte_sstore.unpack_handle(sender, buffer, &ss_handle)) ) {
ORTE_ERROR_LOG(ret);
return;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
options = OBJ_NEW(opal_crs_base_ckpt_options_t);
if( ORTE_SUCCESS != (ret = orte_snapc_base_unpack_options(buffer, options)) ) {
ORTE_ERROR_LOG(ret);
return;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/* In this case we want to use the current_options that are cached
* so that we do not have to send them every time.
*/
opal_crs_base_copy_options(options, global_snapshot.options);
count = 1;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &(loc_migrating), &count, OPAL_BOOL))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( loc_migrating ) {
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &loc_num_procs, &count, OPAL_SIZE))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
for( i = 0; i < loc_num_procs; ++i ) {
count = 1;
if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &proc, &count, ORTE_NAME))) {
ORTE_ERROR_LOG(ret);
goto cleanup;
}
/* JJH: Update local info as needed */
}
2009-09-22 22:26:12 +04:00
}
}
if( ORTE_SUCCESS != (ret = global_coord_job_state_update(jobid,
job_ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ss_handle,
global_snapshot.options) ) ) {
ORTE_ERROR_LOG(ret);
}
cleanup:
2009-09-22 22:26:12 +04:00
if( NULL != options ) {
OBJ_RELEASE(options);
options = NULL;
}
return;
}
static int snapc_full_establish_snapshot_dir(bool empty_metadata)
{
char **value = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
int idx = 0;
/*********************
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* Contact the Stable Storage Framework to setup the storage directory
*********************/
INC_SEQ_NUM();
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_sstore.request_checkpoint_handle(&(global_snapshot.ss_handle),
orte_snapc_base_snapshot_seq_number,
current_global_jobid);
if( currently_migrating ) {
orte_sstore.set_attr(global_snapshot.ss_handle,
SSTORE_METADATA_GLOBAL_MIGRATING,
"1");
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_sstore.register_handle(global_snapshot.ss_handle);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Save the AMCA parameter used into the metadata file
*/
MCA/base: Add new MCA variable system Features: - Support for an override parameter file (openmpi-mca-param-override.conf). Variable values in this file can not be overridden by any file or environment value. - Support for boolean, unsigned, and unsigned long long variables. - Support for true/false values. - Support for enumerations on integer variables. - Support for MPIT scope, verbosity, and binding. - Support for command line source. - Support for setting variable source via the environment using OMPI_MCA_SOURCE_<var name>=source (either command or file:filename) - Cleaner API. - Support for variable groups (equivalent to MPIT categories). Notes: - Variables must be created with a backing store (char **, int *, or bool *) that must live at least as long as the variable. - Creating a variable with the MCA_BASE_VAR_FLAG_SETTABLE enables the use of mca_base_var_set_value() to change the value. - String values are duplicated when the variable is registered. It is up to the caller to free the original value if necessary. The new value will be freed by the mca_base_var system and must not be freed by the user. - Variables with constant scope may not be settable. - Variable groups (and all associated variables) are deregistered when the component is closed or the component repository item is freed. This prevents a segmentation fault from accessing a variable after its component is unloaded. - After some discussion we decided we should remove the automatic registration of component priority variables. Few component actually made use of this feature. - The enumerator interface was updated to be general enough to handle future uses of the interface. - The code to generate ompi_info output has been moved into the MCA variable system. See mca_base_var_dump(). opal: update core and components to mca_base_var system orte: update core and components to mca_base_var system ompi: update core and components to mca_base_var system This commit also modifies the rmaps framework. The following variables were moved from ppr and lama: rmaps_base_pernode, rmaps_base_n_pernode, rmaps_base_n_persocket. Both lama and ppr create synonyms for these variables. This commit was SVN r28236.
2013-03-28 01:09:41 +04:00
if( 0 > (idx = mca_base_var_find("opal", "mca", "base", "param_file_prefix")) ) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
opal_show_help("help-orte-restart.txt", "amca_param_not_found", true);
}
if( 0 < idx ) {
mca_base_var_get_value (idx, &value, NULL, NULL);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
MCA/base: Add new MCA variable system Features: - Support for an override parameter file (openmpi-mca-param-override.conf). Variable values in this file can not be overridden by any file or environment value. - Support for boolean, unsigned, and unsigned long long variables. - Support for true/false values. - Support for enumerations on integer variables. - Support for MPIT scope, verbosity, and binding. - Support for command line source. - Support for setting variable source via the environment using OMPI_MCA_SOURCE_<var name>=source (either command or file:filename) - Cleaner API. - Support for variable groups (equivalent to MPIT categories). Notes: - Variables must be created with a backing store (char **, int *, or bool *) that must live at least as long as the variable. - Creating a variable with the MCA_BASE_VAR_FLAG_SETTABLE enables the use of mca_base_var_set_value() to change the value. - String values are duplicated when the variable is registered. It is up to the caller to free the original value if necessary. The new value will be freed by the mca_base_var system and must not be freed by the user. - Variables with constant scope may not be settable. - Variable groups (and all associated variables) are deregistered when the component is closed or the component repository item is freed. This prevents a segmentation fault from accessing a variable after its component is unloaded. - After some discussion we decided we should remove the automatic registration of component priority variables. Few component actually made use of this feature. - The enumerator interface was updated to be general enough to handle future uses of the interface. - The code to generate ompi_info output has been moved into the MCA variable system. See mca_base_var_dump(). opal: update core and components to mca_base_var system orte: update core and components to mca_base_var system ompi: update core and components to mca_base_var system This commit also modifies the rmaps framework. The following variables were moved from ppr and lama: rmaps_base_pernode, rmaps_base_n_pernode, rmaps_base_n_persocket. Both lama and ppr create synonyms for these variables. This commit was SVN r28236.
2013-03-28 01:09:41 +04:00
if (*value) {
orte_sstore.set_attr(global_snapshot.ss_handle,
SSTORE_METADATA_GLOBAL_AMCA_PARAM,
*value);
MCA/base: Add new MCA variable system Features: - Support for an override parameter file (openmpi-mca-param-override.conf). Variable values in this file can not be overridden by any file or environment value. - Support for boolean, unsigned, and unsigned long long variables. - Support for true/false values. - Support for enumerations on integer variables. - Support for MPIT scope, verbosity, and binding. - Support for command line source. - Support for setting variable source via the environment using OMPI_MCA_SOURCE_<var name>=source (either command or file:filename) - Cleaner API. - Support for variable groups (equivalent to MPIT categories). Notes: - Variables must be created with a backing store (char **, int *, or bool *) that must live at least as long as the variable. - Creating a variable with the MCA_BASE_VAR_FLAG_SETTABLE enables the use of mca_base_var_set_value() to change the value. - String values are duplicated when the variable is registered. It is up to the caller to free the original value if necessary. The new value will be freed by the mca_base_var system and must not be freed by the user. - Variables with constant scope may not be settable. - Variable groups (and all associated variables) are deregistered when the component is closed or the component repository item is freed. This prevents a segmentation fault from accessing a variable after its component is unloaded. - After some discussion we decided we should remove the automatic registration of component priority variables. Few component actually made use of this feature. - The enumerator interface was updated to be general enough to handle future uses of the interface. - The code to generate ompi_info output has been moved into the MCA variable system. See mca_base_var_dump(). opal: update core and components to mca_base_var system orte: update core and components to mca_base_var system ompi: update core and components to mca_base_var system This commit also modifies the rmaps framework. The following variables were moved from ppr and lama: rmaps_base_pernode, rmaps_base_n_pernode, rmaps_base_n_persocket. Both lama and ppr create synonyms for these variables. This commit was SVN r28236.
2013-03-28 01:09:41 +04:00
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) AMCA Parameter Preserved: %s",
*value));
}
}
/*
* Save the TUNE parameter used into the metadata file
*/
if( 0 > (idx = mca_base_var_find("opal", "mca", "base", "envar_file_prefix")) ) {
opal_show_help("help-orte-restart.txt", "tune_param_not_found", true);
}
if( 0 < idx ) {
mca_base_var_get_value (idx, &value, NULL, NULL);
if (*value) {
orte_sstore.set_attr(global_snapshot.ss_handle,
SSTORE_METADATA_GLOBAL_TUNE_PARAM,
*value);
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) TUNE Parameter Preserved: %s",
*value));
}
}
return ORTE_SUCCESS;
}
2009-09-22 22:26:12 +04:00
static int snapc_full_global_checkpoint(opal_crs_base_ckpt_options_t *options)
{
int ret, exit_status = ORTE_SUCCESS;
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) Checkpoint of job %s has been requested\n",
ORTE_JOBID_PRINT(current_global_jobid)));
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/* opal_output(0, "================> JJH Checkpoint Started"); */
2009-09-22 22:26:12 +04:00
current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_REQUEST;
/*********************
* Generate the global snapshot directory, and unique global snapshot handle
*********************/
if( ORTE_SUCCESS != (ret = snapc_full_establish_snapshot_dir(false))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/***********************************
* Do an update handshake with the orte_checkpoint command
***********************************/
updated_job_to_running = false;
2009-09-22 22:26:12 +04:00
if( is_orte_checkpoint_connected &&
ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
current_job_ckpt_state) ) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/**********************
* Notify the Local Snapshot Coordinators of the checkpoint request
**********************/
OPAL_OUTPUT_VERBOSE((15, mca_snapc_full_component.super.output_handle,
"Global) Notifying the Local Coordinators\n"));
2009-09-22 22:26:12 +04:00
if( ORTE_SUCCESS != (ret = snapc_full_global_notify_checkpoint(current_global_jobid, options)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
cleanup:
return exit_status;
}
2009-09-22 22:26:12 +04:00
static int snapc_full_global_notify_checkpoint(orte_jobid_t jobid,
opal_crs_base_ckpt_options_t *options)
{
int ret, exit_status = ORTE_SUCCESS;
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
opal_list_item_t* item = NULL;
int ckpt_state;
2009-09-22 22:26:12 +04:00
ckpt_state = ORTE_SNAPC_CKPT_STATE_PENDING;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Copy over the options
*/
opal_crs_base_copy_options(options, global_snapshot.options);
/*
* Update the global structure
*/
for(item = opal_list_get_first(&global_snapshot.local_snapshots);
item != opal_list_get_end(&global_snapshot.local_snapshots);
item = opal_list_get_next(item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orted_snapshot->state = ckpt_state;
}
/*
* Update the job state, and broadcast to all local daemons
*/
if( ORTE_SUCCESS != (ret = orte_snapc_full_global_set_job_ckpt_info(jobid,
ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
false, options) ) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
cleanup:
return exit_status;
}
/**********************************
* Job/Proc State Set/Get Routines
**********************************/
static int orte_snapc_full_global_set_job_ckpt_info( orte_jobid_t jobid,
int ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_sstore_base_handle_t handle,
2009-09-22 22:26:12 +04:00
bool quick,
opal_crs_base_ckpt_options_t *options)
{
int ret, exit_status = ORTE_SUCCESS;
orte_snapc_full_cmd_flag_t command;
opal_buffer_t *buffer = NULL;
char * state_str = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_proc_t *proc = NULL;
opal_list_item_t *item = NULL;
size_t num_procs;
orte_grpcomm_signature_t *sig;
/*
* Update all Local Coordinators (broadcast operation)
*/
buffer = OBJ_NEW(opal_buffer_t);
if( quick ) {
command = ORTE_SNAPC_FULL_UPDATE_JOB_STATE_QUICK_CMD;
} else {
command = ORTE_SNAPC_FULL_UPDATE_JOB_STATE_CMD;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &jobid, 1, ORTE_JOBID))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &ckpt_state, 1, OPAL_INT))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
if( quick ) {
goto process_msg;
}
if (ORTE_SUCCESS != (ret = orte_sstore.pack_handle(NULL, buffer, handle))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
if(ORTE_SUCCESS != (ret = orte_snapc_base_pack_options(buffer, options))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &(currently_migrating), 1, OPAL_BOOL))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( currently_migrating ) {
num_procs = opal_list_get_size(migrating_procs);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &num_procs, 1, OPAL_SIZE))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
for (item = opal_list_get_first(migrating_procs);
item != opal_list_get_end(migrating_procs);
item = opal_list_get_next(item)) {
proc = (orte_proc_t*)item;
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &(proc->name), 1, ORTE_NAME))) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
2009-09-22 22:26:12 +04:00
}
process_msg:
orte_snapc_ckpt_state_str(&state_str, ckpt_state);
OPAL_OUTPUT_VERBOSE((15, mca_snapc_full_component.super.output_handle,
"Global) Notify Local Coordinators of job %s state change to %d (%s)\n",
ORTE_JOBID_PRINT(jobid), (int)ckpt_state, state_str ));
free(state_str);
state_str = NULL;
/* goes to all daemons */
sig = OBJ_NEW(orte_grpcomm_signature_t);
sig->signature = (orte_process_name_t*)malloc(sizeof(orte_process_name_t));
sig->signature[0].jobid = ORTE_PROC_MY_NAME->jobid;
sig->signature[0].vpid = ORTE_VPID_WILDCARD;
if (ORTE_SUCCESS != (ret = orte_grpcomm.xcast(sig, ORTE_RML_TAG_SNAPC_FULL, buffer))) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
/*
* We will also receive the job update, and process in the RML callback
*/
cleanup:
if( NULL != state_str ) {
free(state_str);
state_str = NULL;
}
OBJ_RELEASE(buffer);
OBJ_RELEASE(sig);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
return exit_status;
}
int global_coord_job_state_update(orte_jobid_t jobid,
int job_ckpt_state,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
orte_sstore_base_handle_t ss_handle,
2009-09-22 22:26:12 +04:00
opal_crs_base_ckpt_options_t *options)
{
int ret, exit_status = ORTE_SUCCESS;
char * state_str = NULL;
orte_snapc_ckpt_state_str(&state_str, job_ckpt_state);
OPAL_OUTPUT_VERBOSE((15, mca_snapc_full_component.super.output_handle,
"Global) Job update command: jobid %s -> state %d (%s)\n",
ORTE_JOBID_PRINT(jobid), (int)job_ckpt_state, state_str ));
free(state_str);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
state_str = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/************************
* Update the orte_checkpoint command
************************/
current_job_ckpt_state = job_ckpt_state;
if( is_orte_checkpoint_connected &&
ORTE_SUCCESS != (ret = orte_snapc_base_global_coord_ckpt_update_cmd(&orte_checkpoint_sender,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
global_snapshot.ss_handle,
current_job_ckpt_state)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
2009-09-22 22:26:12 +04:00
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Global Coordinator: If also a Local coordinator then act locally before globally
*/
if( ORTE_SNAPC_LOCAL_COORD_TYPE == (orte_snapc_coord_type & ORTE_SNAPC_LOCAL_COORD_TYPE) ) {
if( ORTE_SUCCESS != (ret = local_coord_job_state_update(jobid,
job_ckpt_state,
ss_handle,
options)) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
goto cleanup;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Process the cmd
*/
if(ORTE_SNAPC_CKPT_STATE_ESTABLISHED == job_ckpt_state ) {
/*
* If the processes recovered before the checkpoint was established,
* then we need to cleanup here instead of in the recovery block
*/
if( cleanup_on_establish ) {
if( ORTE_SUCCESS != (ret = orte_snapc_full_global_reset_coord()) ) {
ORTE_ERROR_LOG(ret);
exit_status = ret;
goto cleanup;
}
}
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
else if(ORTE_SNAPC_CKPT_STATE_ERROR == job_ckpt_state ) {
opal_output(mca_snapc_full_component.super.output_handle,
"Error: Checkpoint failed!");
}
2009-09-22 22:26:12 +04:00
/*
* This should not happen, since this state is always handled locally
*/
else if(ORTE_SNAPC_CKPT_STATE_STOPPED == job_ckpt_state ) {
;
}
/*
* This should not happen, since we do not handle this case
*/
else if(ORTE_SNAPC_CKPT_STATE_REQUEST == job_ckpt_state ) {
opal_output(mca_snapc_full_component.super.output_handle,
"ERROR: Internal Checkpoint request not implemented.");
ORTE_ERROR_LOG(ORTE_ERR_NOT_SUPPORTED);
}
cleanup:
if( NULL != state_str) {
free(state_str);
state_str = NULL;
}
return exit_status;
}
static int write_out_global_metadata(void)
{
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
opal_list_item_t* orted_item = NULL;
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
"Global) Updating Metadata"));
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Check for an error
* JJH CLEANUP: Check might be good, but mostly unnecessary
* JJH: Do we want to pass this along to the SStore? Probably
*/
for(orted_item = opal_list_get_first(&(global_snapshot.local_snapshots));
orted_item != opal_list_get_end(&(global_snapshot.local_snapshots));
orted_item = opal_list_get_next(orted_item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)orted_item;
if( ORTE_SNAPC_CKPT_STATE_ERROR == orted_snapshot->state ) {
return ORTE_ERROR;
}
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
/*
* Sync the stable storage
*/
orte_sstore.sync(global_snapshot.ss_handle);
SNAPC_FULL_SET_TIMER(SNAPC_FULL_TIMER_SS_SYNC);
return ORTE_SUCCESS;
}
static orte_snapc_full_orted_snapshot_t *find_orted_snapshot(orte_process_name_t *name )
{
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
int ret;
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
opal_list_item_t* item = NULL;
orte_ns_cmp_bitmask_t mask;
for(item = opal_list_get_first(&(global_snapshot.local_snapshots));
item != opal_list_get_end(&(global_snapshot.local_snapshots));
item = opal_list_get_next(item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
mask = ORTE_NS_CMP_ALL;
if (OPAL_EQUAL ==
orte_util_compare_name_fields(mask, name, &orted_snapshot->process_name)) {
return orted_snapshot;
}
}
/*
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
* Refresh the job structure, and try again
*/
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) find_orted(%s) failed. Refreshing and trying again...",
ORTE_NAME_PRINT(name) ));
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( ORTE_SUCCESS != (ret = global_refresh_job_structs()) ) {
ORTE_ERROR_LOG(ret);
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
return NULL;
}
for(item = opal_list_get_first(&(global_snapshot.local_snapshots));
item != opal_list_get_end(&(global_snapshot.local_snapshots));
item = opal_list_get_next(item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
mask = ORTE_NS_CMP_ALL;
if (OPAL_EQUAL ==
orte_util_compare_name_fields(mask, name, &orted_snapshot->process_name)) {
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
return orted_snapshot;
}
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
return NULL;
}
static int snapc_full_global_get_min_state(void)
{
int min_state = ORTE_SNAPC_CKPT_MAX;
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
opal_list_item_t* item = NULL;
char * state_str_a = NULL;
char * state_str_b = NULL;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
current_total_orteds = 0;
for(item = opal_list_get_first(&(global_snapshot.local_snapshots));
item != opal_list_get_end(&(global_snapshot.local_snapshots));
item = opal_list_get_next(item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
2009-09-22 22:26:12 +04:00
/* Ignore orteds with no processes */
if( 0 >= opal_list_get_size(&(orted_snapshot->super.local_snapshots)) ) {
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) ... %s Skipping - (no children)",
2009-09-22 22:26:12 +04:00
ORTE_NAME_PRINT(&orted_snapshot->process_name) ));
continue;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
current_total_orteds++;
if( NULL != state_str_a ) {
free(state_str_a);
state_str_a = NULL;
}
if( NULL != state_str_b ) {
free(state_str_b);
state_str_b = NULL;
}
orte_snapc_ckpt_state_str(&state_str_a, orted_snapshot->state);
orte_snapc_ckpt_state_str(&state_str_b, min_state);
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) ... %s Checking [%d %s] vs [%d %s]",
ORTE_NAME_PRINT(&orted_snapshot->process_name),
(int)orted_snapshot->state, state_str_a,
min_state, state_str_b ));
if( (int)min_state > (int)orted_snapshot->state ) {
min_state = orted_snapshot->state;
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
"Global) ... %s Update --> Min State [%d %s]",
ORTE_NAME_PRINT(&orted_snapshot->process_name),
(int)min_state, state_str_a ));
}
}
if( NULL != state_str_b ) {
free(state_str_b);
state_str_b = NULL;
}
orte_snapc_ckpt_state_str(&state_str_b, min_state);
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
"Global) ... Min State [%d %s]",
(int)min_state, state_str_b ));
if( NULL != state_str_a ) {
free(state_str_a);
state_str_a = NULL;
}
if( NULL != state_str_b ) {
free(state_str_b);
state_str_b = NULL;
}
return min_state;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static int orte_snapc_full_global_reset_coord(void)
{
int ret, exit_status = ORTE_SUCCESS;
opal_list_item_t* item = NULL;
opal_list_item_t* aitem = NULL;
orte_snapc_full_orted_snapshot_t *orted_snapshot = NULL;
orte_snapc_base_local_snapshot_t *app_snapshot = NULL;
/********************************
* Terminate the job if requested
* At this point the application should have already exited, but do this
* just to make doubly sure that the job is terminated.
*********************************/
if( global_snapshot.options->term ) {
SNAPC_FULL_DISPLAY_ALL_TIMERS();
orte_plm.terminate_job(current_global_jobid);
} else {
SNAPC_FULL_DISPLAY_ALL_TIMERS();
}
/*
* Just cleanup, do not need to send out another message
*/
opal_crs_base_clear_options(global_snapshot.options);
/*
* Reset global data structures
*/
for(item = opal_list_get_first(&(global_snapshot.local_snapshots));
item != opal_list_get_end(&(global_snapshot.local_snapshots));
item = opal_list_get_next(item) ) {
orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
orted_snapshot->state = ORTE_SNAPC_CKPT_STATE_NONE;
for(aitem = opal_list_get_first(&(orted_snapshot->super.local_snapshots));
aitem != opal_list_get_end(&(orted_snapshot->super.local_snapshots));
aitem = opal_list_get_next(aitem) ) {
app_snapshot = (orte_snapc_base_local_snapshot_t*)aitem;
app_snapshot->state = ORTE_SNAPC_CKPT_STATE_NONE;
}
}
/************************
* Set up the Command Line listener again
*************************/
is_orte_checkpoint_connected = false;
if( ORTE_SUCCESS != (ret = snapc_full_global_start_cmdline_listener() ) ){
ORTE_ERROR_LOG(ret);
exit_status = ret;
}
current_job_ckpt_state = ORTE_SNAPC_CKPT_STATE_NONE;
cleanup_on_establish = false;
report_progress_cur_loc_finished = 0;
report_progress_last_reported_loc_finished = 0;
return exit_status;
}
/************************
* Timing
************************/
static void snapc_full_set_time(int idx)
{
if(idx < SNAPC_FULL_TIMER_MAX ) {
if( timer_start[idx] <= 0.0 ) {
timer_start[idx] = snapc_full_get_time();
}
}
}
static void snapc_full_display_all_timers(void)
{
double diff = 0.0;
char * label = NULL;
opal_output(0, "Snapshot Coordination Timing: ******************** Summary Begin\n");
/********** Startup time **********/
label = strdup("Running");
diff = timer_start[SNAPC_FULL_TIMER_RUNNING] - timer_start[SNAPC_FULL_TIMER_START];
snapc_full_display_indv_timer_core(diff, label);
free(label);
/********** Time to finish locally **********/
label = strdup("Finish Locally");
diff = timer_start[SNAPC_FULL_TIMER_FIN_LOCAL] - timer_start[SNAPC_FULL_TIMER_RUNNING];
snapc_full_display_indv_timer_core(diff, label);
free(label);
if( timer_start[SNAPC_FULL_TIMER_SS_SYNC] <= timer_start[SNAPC_FULL_TIMER_RECOVERED] ) {
/********** SStore Sync **********/
label = strdup("SStore Sync");
diff = timer_start[SNAPC_FULL_TIMER_SS_SYNC] - timer_start[SNAPC_FULL_TIMER_FIN_LOCAL];
snapc_full_display_indv_timer_core(diff, label);
free(label);
/********** Establish Ckpt **********/
label = strdup("Establish");
diff = timer_start[SNAPC_FULL_TIMER_ESTABLISH] - timer_start[SNAPC_FULL_TIMER_SS_SYNC];
snapc_full_display_indv_timer_core(diff, label);
free(label);
/********** Recover **********/
label = strdup("Continue/Recover");
diff = timer_start[SNAPC_FULL_TIMER_RECOVERED] - timer_start[SNAPC_FULL_TIMER_ESTABLISH];
snapc_full_display_indv_timer_core(diff, label);
free(label);
} else { /* Established after procs recovered */
/********** SStore Sync **********/
label = strdup("SStore Sync*");
diff = timer_start[SNAPC_FULL_TIMER_SS_SYNC] - timer_start[SNAPC_FULL_TIMER_RECOVERED];
snapc_full_display_indv_timer_core(diff, label);
free(label);
/********** Establish Ckpt **********/
label = strdup("Establish*");
diff = timer_start[SNAPC_FULL_TIMER_ESTABLISH] - timer_start[SNAPC_FULL_TIMER_SS_SYNC];
snapc_full_display_indv_timer_core(diff, label);
free(label);
/********** Recover **********/
label = strdup("Continue/Recover*");
diff = timer_start[SNAPC_FULL_TIMER_RECOVERED] - timer_start[SNAPC_FULL_TIMER_FIN_LOCAL];
snapc_full_display_indv_timer_core(diff, label);
free(label);
}
opal_output(0, "Snapshot Coordination Timing: ******************** Summary End\n");
}
static void snapc_full_display_recovered_timers(void)
{
double diff = 0.0;
char * label = NULL;
opal_output(0, "Snapshot Coordination Timing: ******************** Summary Begin\n");
/********** Recover **********/
label = strdup("Recover");
diff = timer_start[SNAPC_FULL_TIMER_RECOVERED] - timer_start[SNAPC_FULL_TIMER_START];
snapc_full_display_indv_timer_core(diff, label);
free(label);
opal_output(0, "Snapshot Coordination Timing: ******************** Summary End\n");
}
static void snapc_full_clear_timers(void)
{
int i;
for(i = 0; i < SNAPC_FULL_TIMER_MAX; ++i) {
timer_start[i] = 0.0;
}
}
static double snapc_full_get_time(void)
{
double wtime;
#if OPAL_TIMER_USEC_NATIVE
wtime = (double)opal_timer_base_get_usec() / 1000000.0;
#else
struct timeval tv;
gettimeofday(&tv, NULL);
wtime = tv.tv_sec;
wtime += (double)tv.tv_usec / 1000000.0;
#endif
return wtime;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
static void snapc_full_display_indv_timer_core(double diff, char *str)
{
double total = 0;
double perc = 0;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( timer_start[SNAPC_FULL_TIMER_SS_SYNC] <= timer_start[SNAPC_FULL_TIMER_RECOVERED] ) {
total = timer_start[SNAPC_FULL_TIMER_RECOVERED] - timer_start[SNAPC_FULL_TIMER_START];
} else {
total = timer_start[SNAPC_FULL_TIMER_ESTABLISH] - timer_start[SNAPC_FULL_TIMER_START];
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
perc = (diff/total) * 100;
opal_output(0,
"snapc_full: timing: %-20s = %10.2f s\t%10.2f s\t%6.2f\n",
str,
diff,
total,
perc);
return;
}
static void snapc_full_report_progress(orte_snapc_full_orted_snapshot_t *orted_snapshot, int total, int min_state)
{
orte_snapc_full_orted_snapshot_t *loc_orted_snapshot = NULL;
opal_list_item_t* item = NULL;
double perc_done;
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL != orted_snapshot->state ) {
return;
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
report_progress_cur_loc_finished++;
perc_done = (total-report_progress_cur_loc_finished)/(total*1.0);
perc_done = (perc_done-1)*(-100.0);
if( perc_done >= (report_progress_last_reported_loc_finished + orte_snapc_full_progress_meter) ||
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
report_progress_last_reported_loc_finished == 0.0 ) {
report_progress_last_reported_loc_finished = perc_done;
opal_output(0, "snapc_full: progress: %10.2f %c Locally Finished\n",
perc_done, '%');
}
A number of C/R enhancements per RFC below: http://www.open-mpi.org/community/lists/devel/2010/07/8240.php Documentation: http://osl.iu.edu/research/ft/ Major Changes: -------------- * Added C/R-enabled Debugging support. Enabled with the --enable-crdebug flag. See the following website for more information: http://osl.iu.edu/research/ft/crdebug/ * Added Stable Storage (SStore) framework for checkpoint storage * 'central' component does a direct to central storage save * 'stage' component stages checkpoints to central storage while the application continues execution. * 'stage' supports offline compression of checkpoints before moving (sstore_stage_compress) * 'stage' supports local caching of checkpoints to improve automatic recovery (sstore_stage_caching) * Added Compression (compress) framework to support * Add two new ErrMgr recovery policies * {{{crmig}}} C/R Process Migration * {{{autor}}} C/R Automatic Recovery * Added the {{{ompi-migrate}}} command line tool to support the {{{crmig}}} ErrMgr component * Added CR MPI Ext functions (enable them with {{{--enable-mpi-ext=cr}}} configure option) * {{{OMPI_CR_Checkpoint}}} (Fixes trac:2342) * {{{OMPI_CR_Restart}}} * {{{OMPI_CR_Migrate}}} (may need some more work for mapping rules) * {{{OMPI_CR_INC_register_callback}}} (Fixes trac:2192) * {{{OMPI_CR_Quiesce_start}}} * {{{OMPI_CR_Quiesce_checkpoint}}} * {{{OMPI_CR_Quiesce_end}}} * {{{OMPI_CR_self_register_checkpoint_callback}}} * {{{OMPI_CR_self_register_restart_callback}}} * {{{OMPI_CR_self_register_continue_callback}}} * The ErrMgr predicted_fault() interface has been changed to take an opal_list_t of ErrMgr defined types. This will allow us to better support a wider range of fault prediction services in the future. * Add a progress meter to: * FileM rsh (filem_rsh_process_meter) * SnapC full (snapc_full_progress_meter) * SStore stage (sstore_stage_progress_meter) * Added 2 new command line options to ompi-restart * --showme : Display the full command line that would have been exec'ed. * --mpirun_opts : Command line options to pass directly to mpirun. (Fixes trac:2413) * Deprecated some MCA params: * crs_base_snapshot_dir deprecated, use sstore_stage_local_snapshot_dir * snapc_base_global_snapshot_dir deprecated, use sstore_base_global_snapshot_dir * snapc_base_global_shared deprecated, use sstore_stage_global_is_shared * snapc_base_store_in_place deprecated, replaced with different components of SStore * snapc_base_global_snapshot_ref deprecated, use sstore_base_global_snapshot_ref * snapc_base_establish_global_snapshot_dir deprecated, never well supported * snapc_full_skip_filem deprecated, use sstore_stage_skip_filem Minor Changes: -------------- * Fixes trac:1924 : {{{ompi-restart}}} now recognizes path prefixed checkpoint handles and does the right thing. * Fixes trac:2097 : {{{ompi-info}}} should now report all available CRS components * Fixes trac:2161 : Manual checkpoint movement. A user can 'mv' a checkpoint directory from the original location to another and still restart from it. * Fixes trac:2208 : Honor various TMPDIR varaibles instead of forcing {{{/tmp}}} * Move {{{ompi_cr_continue_like_restart}}} to {{{orte_cr_continue_like_restart}}} to be more flexible in where this should be set. * opal_crs_base_metadata_write* functions have been moved to SStore to support a wider range of metadata handling functionality. * Cleanup the CRS framework and components to work with the SStore framework. * Cleanup the SnapC framework and components to work with the SStore framework (cleans up these code paths considerably). * Add 'quiesce' hook to CRCP for a future enhancement. * We now require a BLCR version that supports {{{cr_request_file()}}} or {{{cr_request_checkpoint()}}} in order to make the code more maintainable. Note that {{{cr_request_file}}} has been deprecated since 0.7.0, so we prefer to use {{{cr_request_checkpoint()}}}. * Add optional application level INC callbacks (registered through the CR MPI Ext interface). * Increase the {{{opal_cr_thread_sleep_wait}}} parameter to 1000 microseconds to make the C/R thread less aggressive. * {{{opal-restart}}} now looks for cache directories before falling back on stable storage when asked. * {{{opal-restart}}} also support local decompression before restarting * {{{orte-checkpoint}}} now uses the SStore framework to work with the metadata * {{{orte-restart}}} now uses the SStore framework to work with the metadata * Remove the {{{orte-restart}}} preload option. This was removed since the user only needs to select the 'stage' component in order to support this functionality. * Since the '-am' parameter is saved in the metadata, {{{ompi-restart}}} no longer hard codes {{{-am ft-enable-cr}}}. * Fix {{{hnp}}} ErrMgr so that if a previous component in the stack has 'fixed' the problem, then it should be skipped. * Make sure to decrement the number of 'num_local_procs' in the orted when one goes away. * odls now checks the SStore framework to see if it needs to load any checkpoint files before launching (to support 'stage'). This separates the SStore logic from the --preload-[binary|files] options. * Add unique IDs to the named pipes established between the orted and the app in SnapC. This is to better support migration and automatic recovery activities. * Improve the checks for 'already checkpointing' error path. * A a recovery output timer, to show how long it takes to restart a job * Do a better job of cleaning up the old session directory on restart. * Add a local module to the autor and crmig ErrMgr components. These small modules prevent the 'orted' component from attempting a local recovery (Which does not work for MPI apps at the moment) * Add a fix for bounding the checkpointable region between MPI_Init and MPI_Finalize. This commit was SVN r23587. The following Trac tickets were found above: Ticket 1924 --> https://svn.open-mpi.org/trac/ompi/ticket/1924 Ticket 2097 --> https://svn.open-mpi.org/trac/ompi/ticket/2097 Ticket 2161 --> https://svn.open-mpi.org/trac/ompi/ticket/2161 Ticket 2192 --> https://svn.open-mpi.org/trac/ompi/ticket/2192 Ticket 2208 --> https://svn.open-mpi.org/trac/ompi/ticket/2208 Ticket 2342 --> https://svn.open-mpi.org/trac/ompi/ticket/2342 Ticket 2413 --> https://svn.open-mpi.org/trac/ompi/ticket/2413
2010-08-11 00:51:11 +04:00
if( perc_done > 95.0 ) {
opal_output(0, "snapc_full: progress: Waiting on the following daemons (%10.2f %c):", perc_done, '%');
for(item = opal_list_get_first(&(global_snapshot.local_snapshots));
item != opal_list_get_end(&(global_snapshot.local_snapshots));
item = opal_list_get_next(item) ) {
loc_orted_snapshot = (orte_snapc_full_orted_snapshot_t*)item;
if( ORTE_SNAPC_CKPT_STATE_FINISHED_LOCAL != loc_orted_snapshot->state ) {
opal_output(0, "snapc_full: progress: Daemon %s",
ORTE_NAME_PRINT(&loc_orted_snapshot->process_name));
}
}
}
return;
}