a200e4f865
*** THIS RFC INCLUDES A MINOR CHANGE TO THE MPI-RTE INTERFACE *** Note: during the course of this work, it was necessary to completely separate the MPI and RTE progress engines. There were multiple places in the MPI layer where ORTE_WAIT_FOR_COMPLETION was being used. A new OMPI_WAIT_FOR_COMPLETION macro was created (defined in ompi/mca/rte/rte.h) that simply cycles across opal_progress until the provided flag becomes false. Places where the MPI layer blocked waiting for RTE to complete an event have been modified to use this macro. *************************************************************************************** I am reissuing this RFC because of the time that has passed since its original release. Since its initial release and review, I have debugged it further to ensure it fully supports tests like loop_spawn. It therefore seems ready for merge back to the trunk. Given its prior review, I have set the timeout for one week. The code is in https://bitbucket.org/rhc/ompi-oob2 WHAT: Rewrite of ORTE OOB WHY: Support asynchronous progress and a host of other features WHEN: Wed, August 21 SYNOPSIS: The current OOB has served us well, but a number of limitations have been identified over the years. Specifically: * it is only progressed when called via opal_progress, which can lead to hangs or recursive calls into libevent (which is not supported by that code) * we've had issues when multiple NICs are available as the code doesn't "shift" messages between transports - thus, all nodes had to be available via the same TCP interface. * the OOB "unloads" incoming opal_buffer_t objects during the transmission, thus preventing use of OBJ_RETAIN in the code when repeatedly sending the same message to multiple recipients * there is no failover mechanism across NICs - if the selected NIC (or its attached switch) fails, we are forced to abort * only one transport (i.e., component) can be "active" The revised OOB resolves these problems: * async progress is used for all application processes, with the progress thread blocking in the event library * each available TCP NIC is supported by its own TCP module. The ability to asynchronously progress each module independently is provided, but not enabled by default (a runtime MCA parameter turns it "on") * multi-address TCP NICs (e.g., a NIC with both an IPv4 and IPv6 address, or with virtual interfaces) are supported - reachability is determined by comparing the contact info for a peer against all addresses within the range covered by the address/mask pairs for the NIC. * a message that arrives on one TCP NIC is automatically shifted to whatever NIC that is connected to the next "hop" if that peer cannot be reached by the incoming NIC. If no TCP module will reach the peer, then the OOB attempts to send the message via all other available components - if none can reach the peer, then an "error" is reported back to the RML, which then calls the errmgr for instructions. * opal_buffer_t now conforms to standard object rules re OBJ_RETAIN as we no longer "unload" the incoming object * NIC failure is reported to the TCP component, which then tries to resend the message across any other available TCP NIC. If that doesn't work, then the message is given back to the OOB base to try using other components. If all that fails, then the error is reported to the RML, which reports to the errmgr for instructions * obviously from the above, multiple OOB components (e.g., TCP and UD) can be active in parallel * the matching code has been moved to the RML (and out of the OOB/TCP component) so it is independent of transport * routing is done by the individual OOB modules (as opposed to the RML). Thus, both routed and non-routed transports can simultaneously be active * all blocking send/recv APIs have been removed. Everything operates asynchronously. KNOWN LIMITATIONS: * although provision is made for component failover as described above, the code for doing so has not been fully implemented yet. At the moment, if all connections for a given peer fail, the errmgr is notified of a "lost connection", which by default results in termination of the job if it was a lifeline * the IPv6 code is present and compiles, but is not complete. Since the current IPv6 support in the OOB doesn't work anyway, I don't consider this a blocker * routing is performed at the individual module level, yet the active routed component is selected on a global basis. We probably should update that to reflect that different transports may need/choose to route in different ways * obviously, not every error path has been tested nor necessarily covered * determining abnormal termination is more challenging than in the old code as we now potentially have multiple ways of connecting to a process. Ideally, we would declare "connection failed" when *all* transports can no longer reach the process, but that requires some additional (possibly complex) code. For now, the code replicates the old behavior only somewhat modified - i.e., if a module sees its connection fail, it checks to see if it is a lifeline. If so, it notifies the errmgr that the lifeline is lost - otherwise, it notifies the errmgr that a non-lifeline connection was lost. * reachability is determined solely on the basis of a shared subnet address/mask - more sophisticated algorithms (e.g., the one used in the tcp btl) are required to handle routing via gateways * the RML needs to assign sequence numbers to each message on a per-peer basis. The receiving RML will then deliver messages in order, thus preventing out-of-order messaging in the case where messages travel across different transports or a message needs to be redirected/resent due to failure of a NIC This commit was SVN r29058.
1676 строки
58 KiB
C
1676 строки
58 KiB
C
/*
|
|
* Copyright (c) 2004-2012 The Trustees of Indiana University.
|
|
* All rights reserved.
|
|
* Copyright (c) 2004-2005 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.
|
|
* $COPYRIGHT$
|
|
*
|
|
* Additional copyrights may follow
|
|
*
|
|
* $HEADER$
|
|
*/
|
|
|
|
#include "orte_config.h"
|
|
|
|
#include <errno.h>
|
|
#include <sys/types.h>
|
|
#ifdef HAVE_UNISTD_H
|
|
#include <unistd.h>
|
|
#endif /* HAVE_UNISTD_H */
|
|
#ifdef HAVE_FCNTL_H
|
|
#include <fcntl.h>
|
|
#endif /* HAVE_FCNTL_H */
|
|
#ifdef HAVE_SYS_TYPES_H
|
|
#include <sys/types.h>
|
|
#endif /* HAVE_SYS_TYPES_H */
|
|
#ifdef HAVE_SYS_STAT_H
|
|
#include <sys/stat.h> /* for mkfifo */
|
|
#endif /* HAVE_SYS_STAT_H */
|
|
#ifdef HAVE_SIGNAL_H
|
|
#include <signal.h>
|
|
#endif
|
|
#ifdef HAVE_STRING_H
|
|
#include <string.h>
|
|
#endif
|
|
|
|
#include "orte/runtime/orte_cr.h"
|
|
#include "orte/runtime/orte_globals.h"
|
|
#include "orte/runtime/orte_wait.h"
|
|
#include "opal/runtime/opal_cr.h"
|
|
#include "opal/util/output.h"
|
|
#include "opal/mca/event/event.h"
|
|
#include "opal/util/opal_environ.h"
|
|
#include "opal/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/mca/snapc/snapc.h"
|
|
#include "orte/mca/snapc/base/base.h"
|
|
#include "orte/mca/errmgr/errmgr.h"
|
|
#include "orte/mca/grpcomm/grpcomm.h"
|
|
#include "orte/mca/rml/rml.h"
|
|
#include "orte/mca/rml/rml_types.h"
|
|
#include "orte/mca/routed/routed.h"
|
|
#include "orte/mca/routed/base/base.h"
|
|
|
|
#include "snapc_full.h"
|
|
|
|
/************************************
|
|
* Locally Global vars & functions :)
|
|
************************************/
|
|
static void snapc_full_app_signal_handler (int signo);
|
|
static int snapc_full_app_notify_response(opal_cr_ckpt_cmd_state_t resp);
|
|
static int app_notify_resp_stage_1(opal_cr_ckpt_cmd_state_t resp);
|
|
static int app_notify_resp_stage_2(int cr_state );
|
|
static int app_notify_resp_stage_3(int cr_state, bool skip_fin_msg);
|
|
static int app_define_pipe_names(void);
|
|
static int snapc_full_app_notify_reopen_files(void);
|
|
static int snapc_full_app_ckpt_handshake_start(opal_cr_ckpt_cmd_state_t resp);
|
|
static int snapc_full_app_ckpt_handshake_end(int cr_state);
|
|
|
|
static int snapc_full_app_ft_event_update_process_info(orte_process_name_t proc, pid_t pid);
|
|
static int snapc_full_app_finished_msg(int cr_state);
|
|
|
|
static int app_notify_resp_inc_prep_only(int cr_state);
|
|
|
|
static char *app_comm_pipe_r = NULL;
|
|
static char *app_comm_pipe_w = NULL;
|
|
static int app_comm_pipe_r_fd = -1;
|
|
static int app_comm_pipe_w_fd = -1;
|
|
|
|
static opal_crs_base_snapshot_t *local_snapshot = NULL;
|
|
|
|
static bool app_notif_processed = false;
|
|
|
|
static bool currently_migrating = false;
|
|
static bool currently_all_migrating = false;
|
|
|
|
static bool currently_checkpointing = false;
|
|
static int current_unique_id = 0;
|
|
|
|
static int current_cr_state = OPAL_CRS_NONE;
|
|
|
|
static orte_sstore_base_handle_t current_ss_handle = ORTE_SSTORE_HANDLE_INVALID, last_ss_handle = ORTE_SSTORE_HANDLE_INVALID;
|
|
static opal_crs_base_ckpt_options_t *current_options = NULL;
|
|
|
|
/************************
|
|
* Function Definitions
|
|
************************/
|
|
|
|
int app_coord_init()
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
opal_cr_notify_callback_fn_t prev_notify_func;
|
|
orte_snapc_full_cmd_flag_t command = ORTE_SNAPC_FULL_REQUEST_OP_CMD;
|
|
orte_snapc_base_request_op_event_t op_event = ORTE_SNAPC_OP_INIT;
|
|
opal_buffer_t buffer;
|
|
orte_grpcomm_collective_t *coll;
|
|
|
|
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
|
|
"App) Initalized for Application %s\n",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
|
|
|
|
/*
|
|
* Register the INC notification callback
|
|
*/
|
|
opal_cr_reg_notify_callback(snapc_full_app_notify_response, &prev_notify_func);
|
|
|
|
/*
|
|
* Set the pipe names
|
|
*/
|
|
current_unique_id = 0;
|
|
app_define_pipe_names();
|
|
|
|
/*
|
|
* Setup a signal handler to catch and start the proper thread
|
|
* to handle the checkpoint
|
|
*/
|
|
if( SIG_ERR == signal(opal_cr_entry_point_signal, snapc_full_app_signal_handler) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) init: Error: Failed to register signal %d\n",
|
|
opal_cr_entry_point_signal);
|
|
ORTE_ERROR_LOG(OPAL_ERROR);
|
|
exit_status = OPAL_ERROR;
|
|
goto cleanup;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((15, mca_snapc_full_component.super.output_handle,
|
|
"app) Named Pipes (%s) (%s), Signal (%d)",
|
|
app_comm_pipe_r, app_comm_pipe_w, opal_cr_entry_point_signal));
|
|
|
|
/*
|
|
* All processes must sync here, so the Global coordinator can know that
|
|
* it is safe to checkpoint now.
|
|
* Rank 0: Sends confirmation message to the Global Coordinator
|
|
*/
|
|
if( 0 == ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Startup Barrier..."));
|
|
}
|
|
|
|
coll = OBJ_NEW(orte_grpcomm_collective_t);
|
|
coll->id = orte_process_info.peer_init_barrier;
|
|
if( ORTE_SUCCESS != (ret = orte_grpcomm.barrier(coll)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
coll->active = true;
|
|
ORTE_WAIT_FOR_COMPLETION(coll->active);
|
|
OBJ_RELEASE(coll);
|
|
|
|
if( 0 == ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Startup Barrier: Send INIT to HNP...!"));
|
|
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
return ORTE_ERROR;
|
|
}
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(ORTE_PROC_MY_NAME->jobid), 1, ORTE_JOBID))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
return ORTE_ERROR;
|
|
}
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(op_event), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
if (0 > (ret = orte_rml.send_buffer(ORTE_PROC_MY_HNP, &buffer, ORTE_RML_TAG_SNAPC_FULL, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
return ORTE_ERROR;
|
|
}
|
|
|
|
OBJ_DESTRUCT(&buffer);
|
|
}
|
|
|
|
if( 0 == ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Startup Barrier: Done!"));
|
|
}
|
|
|
|
cleanup:
|
|
return exit_status;
|
|
}
|
|
|
|
int app_coord_finalize()
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
orte_snapc_full_cmd_flag_t command = ORTE_SNAPC_FULL_REQUEST_OP_CMD;
|
|
orte_snapc_base_request_op_event_t op_event = ORTE_SNAPC_OP_FIN;
|
|
opal_buffer_t buffer;
|
|
orte_std_cntr_t count;
|
|
orte_grpcomm_collective_t *coll;
|
|
|
|
/*
|
|
* All processes must sync here, so the Global coordinator can know that
|
|
* it is no longer safe to checkpoint.
|
|
* Rank 0: Sends confirmation message to the Global Coordinator
|
|
*/
|
|
if( 0 == ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Shutdown Barrier..."));
|
|
}
|
|
|
|
coll = OBJ_NEW(orte_grpcomm_collective_t);
|
|
coll->id = orte_process_info.peer_init_barrier;
|
|
if( ORTE_SUCCESS != (ret = orte_grpcomm.barrier(coll)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
coll->active = true;
|
|
ORTE_WAIT_FOR_COMPLETION(coll->active);
|
|
|
|
if( 0 == ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Shutdown Barrier: Send FIN to HNP...!"));
|
|
|
|
/* Tell HNP that we are finalizing */
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(ORTE_PROC_MY_NAME->jobid), 1, ORTE_JOBID))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(op_event), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
if (0 > (ret = orte_rml.send_buffer(ORTE_PROC_MY_HNP, &buffer, ORTE_RML_TAG_SNAPC_FULL, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
OBJ_DESTRUCT(&buffer);
|
|
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Shutdown Barrier: Waiting on FIN_ACK...!"));
|
|
|
|
/* Wait for HNP to tell us that it is ok to finish finalization.
|
|
* We could have been checkpointing just as we entered finalize, so we
|
|
* need to wait until the checkpoint is finished before finishing.
|
|
*/
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
if (0 > (ret = orte_rml.recv_buffer(ORTE_PROC_MY_HNP, &buffer, ORTE_RML_TAG_SNAPC_FULL, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &command, &count, ORTE_SNAPC_FULL_CMD))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &op_event, &count, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Shutdown Barrier: Waiting on barrier...!"));
|
|
}
|
|
|
|
coll->id = orte_process_info.peer_fini_barrier;
|
|
if( ORTE_SUCCESS != (ret = orte_grpcomm.barrier(coll)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
if( 0 == ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((3, mca_snapc_full_component.super.output_handle,
|
|
"app) Shutdown Barrier, Done!"));
|
|
}
|
|
|
|
cleanup:
|
|
/* cleanup */
|
|
OBJ_RELEASE(coll);
|
|
|
|
/*
|
|
* Cleanup named pipes
|
|
*/
|
|
if( NULL != app_comm_pipe_r) {
|
|
free(app_comm_pipe_r);
|
|
app_comm_pipe_r = NULL;
|
|
}
|
|
|
|
if( NULL != app_comm_pipe_w) {
|
|
free(app_comm_pipe_w);
|
|
app_comm_pipe_w = NULL;
|
|
}
|
|
|
|
return ORTE_SUCCESS;
|
|
}
|
|
|
|
/******************
|
|
* Local functions
|
|
******************/
|
|
static void snapc_full_app_signal_handler (int signo)
|
|
{
|
|
if( opal_cr_entry_point_signal != signo ) {
|
|
OPAL_OUTPUT_VERBOSE((1, mca_snapc_full_component.super.output_handle,
|
|
"App) signal_handler: Received unknown signal %d",
|
|
signo));
|
|
/* Not our signal */
|
|
return;
|
|
}
|
|
if( currently_checkpointing ) {
|
|
opal_output(0, "snapc:full:(app) Error: Received a signal to checkpoint, but Already checkpointing. Ignoring request!");
|
|
}
|
|
else {
|
|
currently_checkpointing = true;
|
|
/*
|
|
* Signal thread to start checkpoint handshake
|
|
*/
|
|
opal_cr_checkpoint_request = OPAL_CR_STATUS_REQUESTED;
|
|
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) signal_handler: Receive Checkpoint Request."));
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Respond to an asynchronous checkpoint request
|
|
*/
|
|
int snapc_full_app_notify_response(opal_cr_ckpt_cmd_state_t resp)
|
|
{
|
|
static int cr_state;
|
|
int app_pid;
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
/*
|
|
* Clear the options set
|
|
*/
|
|
if( NULL == current_options ) {
|
|
current_options = OBJ_NEW(opal_crs_base_ckpt_options_t);
|
|
}
|
|
|
|
if( opal_cr_currently_stalled ) {
|
|
goto STAGE_1;
|
|
}
|
|
|
|
/* Default: use the fast way */
|
|
orte_cr_continue_like_restart = false;
|
|
orte_cr_flush_restart_files = true;
|
|
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Stage 1..."));
|
|
if( ORTE_SUCCESS != (ret = app_notify_resp_stage_1(resp) ) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
|
|
cr_state = OPAL_CRS_RUNNING;
|
|
current_cr_state = cr_state;
|
|
|
|
#if OPAL_ENABLE_CRDEBUG == 1
|
|
if( current_options->attach_debugger ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: C/R Debug: Wait for debugger..."));
|
|
MPIR_debug_with_checkpoint = true;
|
|
}
|
|
if( current_options->detach_debugger ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: C/R Debug: Do not wait for debugger..."));
|
|
MPIR_debug_with_checkpoint = false;
|
|
}
|
|
#endif
|
|
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Start checkpoint..."));
|
|
STAGE_1:
|
|
opal_cr_currently_stalled = false;
|
|
|
|
app_pid = getpid();
|
|
if( orte_snapc_full_skip_app ) {
|
|
OPAL_OUTPUT_VERBOSE((2, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Skipping App. (%d)\n",
|
|
app_pid));
|
|
ret = ORTE_SUCCESS;
|
|
cr_state = OPAL_CRS_CONTINUE;
|
|
}
|
|
else {
|
|
/*
|
|
* INC: Prepare stack using the registered coordination routine
|
|
*/
|
|
if(OPAL_SUCCESS != (ret = opal_cr_inc_core_prep() ) ) {
|
|
if( OPAL_EXISTS == ret ) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Stalling the checkpoint progress until state is stable again (PID = %d)\n",
|
|
app_pid));
|
|
opal_cr_currently_stalled = true;
|
|
return exit_status;
|
|
}
|
|
else {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: checkpoint notification failed. %d\n", ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* If this is a quiesce_start operation then we can stop here after calling
|
|
* the INC prep. Need to keep the connection open for the quiesce_end()
|
|
* operation though.
|
|
*/
|
|
if( current_options->inc_prep_only ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: INC Prep Only..."));
|
|
return app_notify_resp_inc_prep_only(cr_state);
|
|
} else {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Normal operation..."));
|
|
}
|
|
|
|
/*
|
|
* INC: Take the checkpoint
|
|
*
|
|
* If migrating, only checkpoint if you are the target process
|
|
* otherwise just continue.
|
|
*/
|
|
if( currently_all_migrating ) {
|
|
orte_cr_continue_like_restart = true;
|
|
orte_cr_flush_restart_files = false;
|
|
}
|
|
if( !currently_migrating && currently_all_migrating ) {
|
|
OPAL_OUTPUT_VERBOSE((2, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Skipping App. (%d) - This process is not migrating \n",
|
|
app_pid));
|
|
ret = ORTE_SUCCESS;
|
|
cr_state = OPAL_CRS_CONTINUE;
|
|
}
|
|
else {
|
|
ret = opal_cr_inc_core_ckpt(app_pid, local_snapshot, current_options, &cr_state);
|
|
}
|
|
current_cr_state = cr_state;
|
|
|
|
/*
|
|
* Tell Local Coordinator that we are done with local checkpoint
|
|
* (only if not restarting, on restart we are not attached to the Local
|
|
* Coordinator. )
|
|
*/
|
|
if( OPAL_CRS_RESTART != cr_state ) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Stage 2..."));
|
|
if( ORTE_SUCCESS != (ret = app_notify_resp_stage_2(cr_state) ) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* INC: Recover stack using the registered coordination routine
|
|
*/
|
|
if( !currently_all_migrating ) {
|
|
if( OPAL_SUCCESS != (ret = opal_cr_inc_core_recover(cr_state)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
}
|
|
/*
|
|
* If this is a migrating target process, then do not recover the stack, but terminate.
|
|
* All non-migrating processes will wait in the recovery until the target processes are
|
|
* restarted on the target nodes.
|
|
*/
|
|
else {
|
|
/*
|
|
* If we are one of the processes migrating, then terminate after checkpointing
|
|
*/
|
|
if( currently_migrating ) {
|
|
if( OPAL_CRS_RESTART != cr_state ) {
|
|
current_options->term = true;
|
|
}
|
|
else {
|
|
if( OPAL_SUCCESS != (ret = opal_cr_inc_core_recover(cr_state)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
}
|
|
}
|
|
/*
|
|
* If we are not one of the processes migrating, then wait for release.
|
|
* Need to act like we are restarting during recovery, since the migrating processes
|
|
* will expect this logic.
|
|
*/
|
|
else {
|
|
if( OPAL_SUCCESS != (ret = opal_cr_inc_core_recover(OPAL_CRS_RESTART)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Don't stall any longer */
|
|
opal_cr_stall_check = false;
|
|
|
|
if(OPAL_CRS_RESTART == cr_state) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Restarting... (%s : %d)\n",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), app_pid));
|
|
|
|
current_options->term = false;
|
|
/* Do not respond to the non-existent command line tool */
|
|
goto ckpt_cleanup;
|
|
}
|
|
else if(cr_state == OPAL_CRS_CONTINUE) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Continuing...(%s : %d)\n",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), app_pid));
|
|
; /* Don't need to do anything here */
|
|
}
|
|
else if(cr_state == OPAL_CRS_TERM ) {
|
|
; /* Don't need to do anything here */
|
|
}
|
|
else {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Unknown cr_state(%d) [%d]",
|
|
cr_state, app_pid));
|
|
}
|
|
|
|
ckpt_cleanup:
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Stage 3..."));
|
|
if( ORTE_SUCCESS != (ret = app_notify_resp_stage_3(cr_state, false) )) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto ckpt_cleanup;
|
|
}
|
|
|
|
if( current_options->term ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: User has asked to terminate the application"));
|
|
/* Wait here for termination.
|
|
* If we call 'exit' then the job will fail in an ugly way, instead just
|
|
* wait for the Global coordinator to terminate us.
|
|
*/
|
|
while(1) {
|
|
opal_progress();
|
|
sleep(1);
|
|
}
|
|
}
|
|
|
|
if( NULL != current_options ) {
|
|
OBJ_RELEASE(current_options);
|
|
current_options = NULL;
|
|
}
|
|
|
|
currently_checkpointing = false;
|
|
|
|
return exit_status;
|
|
}
|
|
|
|
static int app_notify_resp_stage_1(opal_cr_ckpt_cmd_state_t resp)
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
OPAL_CR_CLEAR_TIMERS();
|
|
opal_cr_timing_my_rank = ORTE_PROC_MY_NAME->vpid;
|
|
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_ENTRY0);
|
|
|
|
/*
|
|
* Open communication channels
|
|
*/
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Open Communication Channels."));
|
|
if (ORTE_SUCCESS != (ret = snapc_full_app_notify_reopen_files())) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
/*
|
|
* Initial Handshake
|
|
*/
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Initial Handshake."));
|
|
if( ORTE_SUCCESS != (ret = snapc_full_app_ckpt_handshake_start(resp) ) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_ENTRY1);
|
|
|
|
/*
|
|
* Register with SStore
|
|
*/
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Register with SStore..."));
|
|
if( OPAL_SUCCESS != (ret = orte_sstore.register_handle(current_ss_handle)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
local_snapshot = OBJ_NEW(opal_crs_base_snapshot_t);
|
|
|
|
if( !currently_migrating && currently_all_migrating ) {
|
|
orte_sstore.set_attr(current_ss_handle,
|
|
SSTORE_METADATA_LOCAL_SKIP_CKPT,
|
|
"1");
|
|
}
|
|
|
|
orte_sstore.get_attr(current_ss_handle,
|
|
SSTORE_METADATA_LOCAL_SNAP_LOC,
|
|
&(local_snapshot->snapshot_directory));
|
|
orte_sstore.get_attr(current_ss_handle,
|
|
SSTORE_METADATA_LOCAL_SNAP_META,
|
|
&(local_snapshot->metadata_filename));
|
|
|
|
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_ENTRY2);
|
|
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Start checkpoint... (%d)", (int)current_ss_handle));
|
|
|
|
cleanup:
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Are we migrating [%5s]. Am I migrating [%5s]",
|
|
(currently_all_migrating ? "True" : "False"),
|
|
(currently_migrating ? "True" : "False") ));
|
|
|
|
return exit_status;
|
|
}
|
|
|
|
static int app_notify_resp_inc_prep_only(int cr_state)
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
/*
|
|
* Tell the local coordinator that we are done with the INC prep
|
|
*/
|
|
if( sizeof(int) != (ret = write(app_comm_pipe_w_fd, &cr_state, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to write cr_state to named pipe (%s).\n",
|
|
app_comm_pipe_w);
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
app_notif_processed = true;
|
|
|
|
cleanup:
|
|
return exit_status;
|
|
}
|
|
|
|
static int app_notify_resp_stage_2(int cr_state )
|
|
{
|
|
int ret;
|
|
|
|
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_ENTRY3);
|
|
|
|
/*
|
|
* Sync SStore
|
|
* If we stopped the process, then we already did this
|
|
*/
|
|
if( !(current_options->stop) ) {
|
|
if( currently_migrating || !currently_all_migrating ) {
|
|
orte_sstore.set_attr(current_ss_handle,
|
|
SSTORE_METADATA_LOCAL_CRS_COMP,
|
|
local_snapshot->component_name);
|
|
}
|
|
|
|
orte_sstore.sync(current_ss_handle);
|
|
}
|
|
last_ss_handle = current_ss_handle;
|
|
current_ss_handle = 0;
|
|
|
|
/*
|
|
* Final Handshake
|
|
*/
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Waiting for final handshake."));
|
|
if( ORTE_SUCCESS != (ret = snapc_full_app_ckpt_handshake_end(cr_state ) ) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
return ret;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Final Handshake complete."));
|
|
|
|
return ORTE_SUCCESS;
|
|
}
|
|
|
|
static int app_define_pipe_names(void)
|
|
{
|
|
if( NULL != app_comm_pipe_r ) {
|
|
free(app_comm_pipe_r);
|
|
app_comm_pipe_r = NULL;
|
|
}
|
|
|
|
if( NULL != app_comm_pipe_w ) {
|
|
free(app_comm_pipe_w);
|
|
app_comm_pipe_w = NULL;
|
|
}
|
|
|
|
asprintf(&app_comm_pipe_r, "%s/%s.%d_%d",
|
|
opal_cr_pipe_dir, OPAL_CR_NAMED_PROG_R,
|
|
(int)getpid(), current_unique_id);
|
|
asprintf(&app_comm_pipe_w, "%s/%s.%d_%d",
|
|
opal_cr_pipe_dir, OPAL_CR_NAMED_PROG_W,
|
|
(int)getpid(), current_unique_id);
|
|
|
|
++current_unique_id;
|
|
|
|
return ORTE_SUCCESS;
|
|
}
|
|
|
|
static int app_notify_resp_stage_3(int cr_state, bool skip_fin_msg)
|
|
{
|
|
/*
|
|
* Send a message to the local daemon letting it know that we are done
|
|
*/
|
|
if( !skip_fin_msg ) {
|
|
snapc_full_app_finished_msg(cr_state);
|
|
}
|
|
|
|
/*
|
|
* Close and cleanup pipes
|
|
*/
|
|
if( 0 <= app_comm_pipe_r_fd ) {
|
|
close(app_comm_pipe_r_fd);
|
|
app_comm_pipe_r_fd = -1;
|
|
}
|
|
if( 0 <= app_comm_pipe_w_fd ) {
|
|
close(app_comm_pipe_w_fd);
|
|
app_comm_pipe_w_fd = -1;
|
|
}
|
|
|
|
remove(app_comm_pipe_r);
|
|
remove(app_comm_pipe_w);
|
|
|
|
app_comm_pipe_r_fd = -1;
|
|
app_comm_pipe_w_fd = -1;
|
|
|
|
if( OPAL_CRS_RESTART == cr_state ) {
|
|
current_unique_id = 0;
|
|
}
|
|
|
|
app_define_pipe_names();
|
|
|
|
/* Prepare to wait for another checkpoint action */
|
|
opal_cr_checkpointing_state = OPAL_CR_STATUS_NONE;
|
|
opal_cr_currently_stalled = false;
|
|
|
|
currently_all_migrating = false;
|
|
currently_migrating = false;
|
|
|
|
OPAL_CR_SET_TIMER(OPAL_CR_TIMER_ENTRY4);
|
|
if(OPAL_CRS_RESTART != cr_state) {
|
|
OPAL_CR_DISPLAY_ALL_TIMERS();
|
|
}
|
|
|
|
return ORTE_SUCCESS;
|
|
}
|
|
|
|
static int snapc_full_app_finished_msg(int cr_state) {
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
opal_buffer_t buffer;
|
|
orte_snapc_cmd_flag_t command = ORTE_SNAPC_LOCAL_FINISH_CMD;
|
|
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &command, 1, ORTE_SNAPC_CMD )) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &cr_state, 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
if (0 > (ret = orte_rml.send_buffer(ORTE_PROC_MY_DAEMON, &buffer, ORTE_RML_TAG_SNAPC, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
cleanup:
|
|
OBJ_DESTRUCT(&buffer);
|
|
|
|
return exit_status;
|
|
}
|
|
|
|
static int snapc_full_app_notify_reopen_files(void)
|
|
{
|
|
int ret = OPAL_ERR_NOT_IMPLEMENTED;
|
|
|
|
#ifndef HAVE_MKFIFO
|
|
return ret;
|
|
#else
|
|
/*
|
|
* Open up the read pipe
|
|
*/
|
|
if( (ret = mkfifo(app_comm_pipe_r, 0660)) < 0) {
|
|
if(EEXIST == ret || -1 == ret ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_reopen_files: mkfifo failed because file (%s) already exists, attempting to use this pipe. (%d)",
|
|
app_comm_pipe_r, ret));
|
|
}
|
|
else {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_reopen_files: Error: mkfifo failed to make named pipe (%s). (%d)\n",
|
|
app_comm_pipe_r, ret);
|
|
return ORTE_ERROR;
|
|
}
|
|
}
|
|
|
|
app_comm_pipe_r_fd = open(app_comm_pipe_r, O_RDWR);
|
|
if(app_comm_pipe_r_fd < 0) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) init: Error: open failed to open the named pipe (%s). %d\n",
|
|
app_comm_pipe_r, app_comm_pipe_r_fd);
|
|
return ORTE_ERROR;
|
|
}
|
|
|
|
/*
|
|
* Open up the write pipe
|
|
*/
|
|
if( (ret = mkfifo(app_comm_pipe_w, 0660)) < 0) {
|
|
if(EEXIST == ret || -1 == ret ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_reopen_files: mkfifo failed because file (%s) already exists, attempting to use this pipe. (%d)",
|
|
app_comm_pipe_w, ret));
|
|
}
|
|
else {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_reopen_files: Error: mkfifo failed to make named pipe (%s). (%d)\n",
|
|
app_comm_pipe_w, ret);
|
|
return ORTE_ERROR;
|
|
}
|
|
}
|
|
|
|
app_comm_pipe_w_fd = open(app_comm_pipe_w, O_WRONLY);
|
|
if(app_comm_pipe_w_fd < 0) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_reopen_files: Error: open failed to open the named pipe (%s). (%d)\n",
|
|
app_comm_pipe_w, app_comm_pipe_w_fd);
|
|
return ORTE_ERROR;
|
|
}
|
|
|
|
return ORTE_SUCCESS;
|
|
#endif /* HAVE_MKFIFO */
|
|
}
|
|
|
|
static int snapc_full_app_ckpt_handshake_start(opal_cr_ckpt_cmd_state_t resp)
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
int tmp_resp, opt_rep;
|
|
|
|
/*
|
|
* Get the initial handshake command:
|
|
* - Migrating option [all, me]
|
|
* - Term argument
|
|
* - Stop argument
|
|
*/
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the all_migrating option from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
currently_all_migrating = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the migrating option from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
currently_migrating = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'term' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
current_options->term = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'stop' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
current_options->stop = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'inc_prep_only' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
current_options->inc_prep_only = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'inc_recover_only' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
current_options->inc_recover_only = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
#if OPAL_ENABLE_CRDEBUG == 1
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'attach_debugger' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
current_options->attach_debugger = OPAL_INT_TO_BOOL(opt_rep);
|
|
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'detach_debugger' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
current_options->detach_debugger = OPAL_INT_TO_BOOL(opt_rep);
|
|
#endif
|
|
|
|
/*
|
|
* Get SStore Handle
|
|
*/
|
|
if( sizeof(orte_sstore_base_handle_t) != (ret = read(app_comm_pipe_r_fd, ¤t_ss_handle, sizeof(orte_sstore_base_handle_t))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the sstore handle from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) %s Received Options... Responding with %d\n",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), (int)resp));
|
|
|
|
/*
|
|
* Write back the response to the request (message printed below)
|
|
*/
|
|
tmp_resp = (int)resp;
|
|
if( sizeof(int) != (ret = write(app_comm_pipe_w_fd, &tmp_resp, sizeof(int)) ) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: %d: Error: Unable to write to pipe (%s) ret = %d [Line %d]\n",
|
|
tmp_resp, app_comm_pipe_w, ret, __LINE__);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
/*
|
|
* Respond that the checkpoint is currently in progress
|
|
*/
|
|
if( OPAL_CHECKPOINT_CMD_IN_PROGRESS == resp ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Checkpoint in progress, cannot start (%d)",
|
|
getpid()));
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
/*
|
|
* Respond that the application is unable to be checkpointed
|
|
*/
|
|
else if( OPAL_CHECKPOINT_CMD_NULL == resp ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Non-checkpointable application, cannot start (%d)",
|
|
getpid()));
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
/*
|
|
* Respond that some error has occurred such that the application is
|
|
* not able to be checkpointed
|
|
*/
|
|
else if( OPAL_CHECKPOINT_CMD_ERROR == resp ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error generated, cannot start (%d)",
|
|
getpid()));
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
/*
|
|
* Respond signalng that we wish to respond to this request
|
|
*/
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Starting checkpoint request (%d)",
|
|
getpid()));
|
|
|
|
/*
|
|
* Get the sentinel value indicating that we can start now
|
|
* JJH: Check for an error here indicating that even though this process is
|
|
* OK to checkpoint others might not be in which case we should cleanup
|
|
* properly.
|
|
*/
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &opt_rep, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
cleanup:
|
|
return exit_status;
|
|
}
|
|
|
|
static int snapc_full_app_ckpt_handshake_end(int cr_state)
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
int last_cmd = 0;
|
|
int err;
|
|
|
|
/*
|
|
* Return the final checkpoint state to the local coordinator
|
|
*/
|
|
if( sizeof(int) != (ret = write(app_comm_pipe_w_fd, &cr_state, sizeof(int))) ) {
|
|
err = errno;
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to write cr_state to named pipe (%s). %d/%d/%s\n",
|
|
app_comm_pipe_w, ret, err, strerror(err));
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
if( currently_all_migrating && currently_migrating ) {
|
|
app_notify_resp_stage_3(cr_state, true);
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) handshake_end: Waiting for termination (%d)",
|
|
getpid()));
|
|
/* Wait here for termination, do not terminate ourselves.
|
|
* JJH: We cannot terminate ourselves without killing the job...
|
|
*/
|
|
while(1) {
|
|
opal_progress();
|
|
sleep(1);
|
|
}
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) handshake_end: Waiting for release (%d)",
|
|
getpid()));
|
|
|
|
/*
|
|
* Wait for the local coordinator to release us
|
|
*/
|
|
if( sizeof(int) != (ret = read(app_comm_pipe_r_fd, &last_cmd, sizeof(int))) ) {
|
|
opal_output(mca_snapc_full_component.super.output_handle,
|
|
"App) notify_response: Error: Unable to read the 'last_cmd' from named pipe (%s). %d\n",
|
|
app_comm_pipe_r, ret);
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) handshake_end: Released... (%d)",
|
|
getpid()));
|
|
|
|
cleanup:
|
|
return exit_status;
|
|
}
|
|
|
|
int app_coord_ft_event(int state) {
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
|
|
"App) In ft_event(%d)", state));
|
|
|
|
/******** Checkpoint Prep ********/
|
|
if(OPAL_CRS_CHECKPOINT == state) {
|
|
/*
|
|
* Record the job session directory
|
|
* This way we will recreate it on restart so that any components that
|
|
* have old references to it (like btl/sm) can reference their files
|
|
* (to close the fd's to them) on restart. We will remove it before we
|
|
* create the new session directory.
|
|
*/
|
|
orte_sstore.set_attr(orte_sstore_handle_current,
|
|
SSTORE_METADATA_LOCAL_MKDIR,
|
|
orte_process_info.job_session_dir);
|
|
|
|
/*
|
|
* If stopping then sync early
|
|
*/
|
|
if( current_options->stop ) {
|
|
orte_sstore.set_attr(current_ss_handle,
|
|
SSTORE_METADATA_LOCAL_CRS_COMP,
|
|
opal_crs_base_selected_component.base_version.mca_component_name);
|
|
|
|
orte_sstore.sync(current_ss_handle);
|
|
}
|
|
}
|
|
/******** Continue Recovery ********/
|
|
else if (OPAL_CRS_CONTINUE == state ) {
|
|
#if OPAL_ENABLE_CRDEBUG == 1
|
|
/*
|
|
* Send PID to HNP/daemon if debugging as an indicator that we have
|
|
* finished the checkpoint operation.
|
|
*/
|
|
if( ORTE_SUCCESS != (ret = snapc_full_app_ft_event_update_process_info(orte_process_info.my_name, getpid())) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
#endif
|
|
; /* Nothing */
|
|
}
|
|
/******** Restart Pre-Recovery ********/
|
|
else if (OPAL_CRS_RESTART_PRE == state ) {
|
|
; /* Nothing */
|
|
}
|
|
/******** Restart Recovery ********/
|
|
else if (OPAL_CRS_RESTART == state ) {
|
|
OPAL_OUTPUT_VERBOSE((20, mca_snapc_full_component.super.output_handle,
|
|
"App) Initalized for Application %s (Restart) (%5d)\n",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), getpid()));
|
|
|
|
/*
|
|
* Send new PID to HNP/daemon
|
|
* The checkpointer could have used a proxy program to boot us
|
|
* so the pid that the orted got from fork() may not be the
|
|
* PID of this application.
|
|
* - Note: BLCR does this because it tries to preseve the PID
|
|
* of the program across checkpointes
|
|
*/
|
|
if( ORTE_SUCCESS != (ret = snapc_full_app_ft_event_update_process_info(orte_process_info.my_name, getpid())) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
/*
|
|
* Since this process is interacting with an 'old' daemon, we must make
|
|
* sure to sync twice.
|
|
* JJH: This assumes that we only move whole nodes, this may be wrong
|
|
* JJH: when interacting with partial migration
|
|
*/
|
|
if( currently_all_migrating && !currently_migrating ) {
|
|
OPAL_OUTPUT_VERBOSE((10, mca_snapc_full_component.super.output_handle,
|
|
"App) ft_event(RESTART): Not a migrating process, so re-sync"));
|
|
orte_routed_base_register_sync(false);
|
|
}
|
|
|
|
/*
|
|
* JJH: Optionally the non-migrating processes can wait here in stage_2
|
|
* JJH: This will delay the initial checkpoint, but potentially speed up
|
|
* JJH: restart.
|
|
*/
|
|
}
|
|
/******** Termination ********/
|
|
else if (OPAL_CRS_TERM == state ) {
|
|
; /* Nothing */
|
|
}
|
|
/******** Error State ********/
|
|
else {
|
|
; /* Nothing */
|
|
}
|
|
|
|
cleanup:
|
|
return exit_status;
|
|
}
|
|
|
|
static int snapc_full_app_ft_event_update_process_info(orte_process_name_t proc, pid_t proc_pid)
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
opal_buffer_t buffer;
|
|
orte_snapc_cmd_flag_t command = ORTE_SNAPC_LOCAL_UPDATE_CMD;
|
|
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &command, 1, ORTE_SNAPC_CMD )) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
/* JJH CLEANUP: Do we really need this, it is equal to sender */
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &proc, 1, ORTE_NAME))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &proc_pid, 1, OPAL_PID))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
#if OPAL_ENABLE_CRDEBUG == 1
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &MPIR_debug_with_checkpoint, 1, OPAL_BOOL))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
#endif
|
|
|
|
if (0 > (ret = orte_rml.send_buffer(ORTE_PROC_MY_DAEMON, &buffer, ORTE_RML_TAG_SNAPC, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
cleanup:
|
|
OBJ_DESTRUCT(&buffer);
|
|
|
|
return exit_status;
|
|
}
|
|
|
|
int app_coord_request_op(orte_snapc_base_request_op_t *datum)
|
|
{
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
orte_snapc_full_cmd_flag_t command = ORTE_SNAPC_FULL_REQUEST_OP_CMD;
|
|
opal_buffer_t buffer;
|
|
orte_std_cntr_t count;
|
|
int op_event, op_state;
|
|
char *seq_str = NULL, *tmp_str = NULL;
|
|
int cr_state = OPAL_CRS_CONTINUE;
|
|
int app_pid, i;
|
|
|
|
/*
|
|
* Quiesce_end recovers the library before talking to the Global coord.
|
|
*/
|
|
if( ORTE_SNAPC_OP_QUIESCE_END == datum->event) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Quiesce_end: Recovering the stack..."));
|
|
|
|
/*
|
|
* INC: Recover the stack
|
|
*/
|
|
if( NULL == local_snapshot->component_name ) {
|
|
local_snapshot->component_name = strdup("");
|
|
}
|
|
if( ORTE_SUCCESS != (ret = app_notify_resp_stage_2(cr_state) ) ) {
|
|
exit_status = ret;
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
if(OPAL_SUCCESS != (ret = opal_cr_inc_core_recover(cr_state) ) ) {
|
|
exit_status = ret;
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
if( ORTE_SUCCESS != (ret = app_notify_resp_stage_3(cr_state, false) )) {
|
|
exit_status = ret;
|
|
ORTE_ERROR_LOG(ret);
|
|
goto cleanup;
|
|
}
|
|
|
|
currently_checkpointing = false;
|
|
app_notif_processed = false;
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Quiesce_end: Recovered."));
|
|
}
|
|
else if( ORTE_SNAPC_OP_QUIESCE_CHECKPOINT == datum->event) {
|
|
app_pid = getpid();
|
|
cr_state = OPAL_CRS_RUNNING;
|
|
if( OPAL_SUCCESS != (ret = opal_cr_inc_core_ckpt(app_pid, local_snapshot, current_options, &cr_state)) ) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
}
|
|
|
|
if( OPAL_CRS_RESTART != cr_state ) {
|
|
orte_sstore.sync(current_ss_handle);
|
|
}
|
|
|
|
orte_sstore.get_attr(current_ss_handle,
|
|
SSTORE_METADATA_GLOBAL_SNAP_SEQ,
|
|
&seq_str);
|
|
if( NULL != seq_str ) {
|
|
datum->seq_num = atoi(seq_str);
|
|
} else {
|
|
datum->seq_num = -1;
|
|
}
|
|
|
|
orte_sstore.get_attr(current_ss_handle,
|
|
SSTORE_METADATA_GLOBAL_SNAP_REF,
|
|
&(datum->global_handle));
|
|
if( NULL == datum->global_handle ) {
|
|
datum->global_handle = strdup("Unknown");
|
|
}
|
|
|
|
return exit_status;
|
|
}
|
|
|
|
/*
|
|
* Leader: Send the info to the head node
|
|
*/
|
|
if( datum->leader == (int)ORTE_PROC_MY_NAME->vpid ) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Request_op: Sending request (%3d)...",
|
|
datum->event));
|
|
/*
|
|
* Send request to HNP
|
|
*/
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &command, 1, ORTE_SNAPC_FULL_CMD))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(ORTE_PROC_MY_NAME->jobid), 1, ORTE_JOBID))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(datum->event), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
if( ORTE_SNAPC_OP_RESTART == datum->event) {
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(datum->seq_num), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(datum->global_handle), 1, OPAL_STRING))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
}
|
|
else if( ORTE_SNAPC_OP_MIGRATE == datum->event) {
|
|
/*
|
|
* Check information
|
|
* Rank | Hostname | cr_off_node | Meaning
|
|
* -------+-----------+--------------+---------
|
|
* self | home/same | false | Do not move this process
|
|
* | | true | ERROR
|
|
* | NULL | false | Move wherever
|
|
* | | true | Move off of this node
|
|
* | other | false/true | Move to the 'other' node
|
|
* -------+-----------+--------------+---------
|
|
* peer | home/same | false | Move 'peer' to me
|
|
* | | true | ERROR
|
|
* | NULL | false | Move wherever (Default: Move 'peer' to me)
|
|
* | | true | Move with peer to some other node
|
|
* | other | false/true | Move with peer to 'other' node
|
|
* -------+-----------+--------------+---------
|
|
* If 'rank' is set to a peer other than self, and the peer sets
|
|
* conflicting 'hostname' or 'cr_off_node' preferences, then that
|
|
* is an error. In which case the migration should fail.
|
|
*/
|
|
currently_all_migrating = true;
|
|
|
|
/*
|
|
* Send information
|
|
*/
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &(datum->mig_num), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
for( i = 0; i < datum->mig_num; ++i ) {
|
|
OPAL_OUTPUT_VERBOSE((30, mca_snapc_full_component.super.output_handle,
|
|
"App) Migration %3d/%3d: Sending 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')
|
|
));
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &((datum->mig_vpids)[i]), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
tmp_str = strdup((datum->mig_host_pref)[i]);
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &tmp_str, 1, OPAL_STRING))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
if( NULL != tmp_str ) {
|
|
free(tmp_str);
|
|
tmp_str = NULL;
|
|
}
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &((datum->mig_vpid_pref)[i]), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&buffer, &((datum->mig_off_node)[i]), 1, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (0 > (ret = orte_rml.send_buffer(ORTE_PROC_MY_HNP, &buffer, ORTE_RML_TAG_SNAPC_FULL, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
OBJ_DESTRUCT(&buffer);
|
|
}
|
|
|
|
/*
|
|
* Wait for the response
|
|
*/
|
|
if( ORTE_SNAPC_OP_CHECKPOINT == datum->event) {
|
|
if( datum->leader == (int)ORTE_PROC_MY_NAME->vpid ) {
|
|
/*
|
|
* Wait for local completion (need to check to see if we are restarting)
|
|
*/
|
|
while(OPAL_CRS_CONTINUE != current_cr_state &&
|
|
OPAL_CRS_RESTART != current_cr_state &&
|
|
OPAL_CRS_ERROR != current_cr_state ) {
|
|
opal_progress();
|
|
OPAL_CR_TEST_CHECKPOINT_READY();
|
|
}
|
|
|
|
/* Do not wait for a response if we are restarting (it will never arrive) */
|
|
if( OPAL_CRS_RESTART == current_cr_state ) {
|
|
orte_sstore.get_attr(current_ss_handle,
|
|
SSTORE_METADATA_GLOBAL_SNAP_SEQ,
|
|
&seq_str);
|
|
if( NULL != seq_str ) {
|
|
datum->seq_num = atoi(seq_str);
|
|
} else {
|
|
datum->seq_num = -1;
|
|
}
|
|
|
|
orte_sstore.get_attr(current_ss_handle,
|
|
SSTORE_METADATA_GLOBAL_SNAP_REF,
|
|
&(datum->global_handle));
|
|
if( NULL == datum->global_handle ) {
|
|
datum->global_handle = strdup("Unknown");
|
|
}
|
|
|
|
current_cr_state = OPAL_CRS_NONE;
|
|
|
|
exit_status = ORTE_SUCCESS;
|
|
goto cleanup;
|
|
}
|
|
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
/*
|
|
* Wait for a response regarding completion
|
|
*/
|
|
if (0 > (ret = orte_rml.recv_buffer(ORTE_PROC_MY_HNP, &buffer, ORTE_RML_TAG_SNAPC_FULL, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &command, &count, ORTE_SNAPC_FULL_CMD))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &op_event, &count, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &op_state, &count, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
OBJ_DESTRUCT(&buffer);
|
|
|
|
orte_sstore.get_attr(last_ss_handle,
|
|
SSTORE_METADATA_GLOBAL_SNAP_SEQ,
|
|
&seq_str);
|
|
datum->seq_num = atoi(seq_str);
|
|
|
|
orte_sstore.get_attr(last_ss_handle,
|
|
SSTORE_METADATA_GLOBAL_SNAP_REF,
|
|
&(datum->global_handle));
|
|
}
|
|
}
|
|
/*
|
|
* Restart will terminate this process, so just wait...
|
|
*/
|
|
else if( ORTE_SNAPC_OP_RESTART == datum->event) {
|
|
while( 1 ) {
|
|
opal_progress();
|
|
OPAL_CR_TEST_CHECKPOINT_READY();
|
|
sleep(1);
|
|
}
|
|
}
|
|
/*
|
|
* Leader waits for response
|
|
*/
|
|
else if( ORTE_SNAPC_OP_MIGRATE == datum->event) {
|
|
if( datum->leader == (int)ORTE_PROC_MY_NAME->vpid ) {
|
|
while( currently_all_migrating ) {
|
|
opal_progress();
|
|
OPAL_CR_TEST_CHECKPOINT_READY();
|
|
sleep(1);
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Request_op: Leader waiting for Migrate release (%3d)...",
|
|
datum->event));
|
|
|
|
OBJ_CONSTRUCT(&buffer, opal_buffer_t);
|
|
|
|
/*
|
|
* Wait for a response regarding completion
|
|
*/
|
|
if (0 > (ret = orte_rml.recv_buffer(ORTE_PROC_MY_HNP, &buffer, ORTE_RML_TAG_SNAPC_FULL, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
OBJ_DESTRUCT(&buffer);
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &command, &count, ORTE_SNAPC_FULL_CMD))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &op_event, &count, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (ret = opal_dss.unpack(&buffer, &op_state, &count, OPAL_INT))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
exit_status = ret;
|
|
goto cleanup;
|
|
}
|
|
|
|
OBJ_DESTRUCT(&buffer);
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Request_op: Leader continuing from Migration (%3d)...",
|
|
datum->event));
|
|
}
|
|
}
|
|
/*
|
|
* Everyone waits here for completion of Quiesce start
|
|
*/
|
|
else if( ORTE_SNAPC_OP_QUIESCE_START == datum->event) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Quiesce_start: Waiting for release..."));
|
|
|
|
while( !app_notif_processed ) {
|
|
opal_progress();
|
|
OPAL_CR_TEST_CHECKPOINT_READY();
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Quiesce_start: Released"));
|
|
}
|
|
/*
|
|
* No waiting for Quiesce end (barrier occurs in protocol)
|
|
*/
|
|
else if( ORTE_SNAPC_OP_QUIESCE_END == datum->event) {
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Quiesce_end: Waiting for release..."));
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, mca_snapc_full_component.super.output_handle,
|
|
"App) Quiesce_end: Released"));
|
|
}
|
|
|
|
|
|
cleanup:
|
|
if( NULL != seq_str ) {
|
|
free(seq_str);
|
|
seq_str = NULL;
|
|
}
|
|
|
|
if( NULL != tmp_str ) {
|
|
free(tmp_str);
|
|
tmp_str = NULL;
|
|
}
|
|
|
|
return exit_status;
|
|
}
|