2007-03-17 02:11:45 +03:00
|
|
|
/*
|
2009-04-30 20:55:39 +04:00
|
|
|
* Copyright (c) 2004-2009 The Trustees of Indiana University and Indiana
|
2007-03-17 02:11:45 +03:00
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2005 The University of Tennessee and The University
|
|
|
|
* of Tennessee Research Foundation. 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.
|
2007-06-05 07:03:59 +04:00
|
|
|
* Copyright (c) 2007 Los Alamos National Security, LLC. All rights
|
|
|
|
* reserved.
|
2007-03-17 02:11:45 +03:00
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* @file
|
|
|
|
* ORTE Checkpoint Tool for checkpointing a multiprocess job
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "orte_config.h"
|
2008-02-28 04:57:57 +03:00
|
|
|
#include "orte/constants.h"
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <errno.h>
|
|
|
|
#ifdef HAVE_STDLIB_H
|
|
|
|
#include <stdlib.h>
|
|
|
|
#endif /* HAVE_STDLIB_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_SYS_WAIT_H
|
|
|
|
#include <sys/wait.h>
|
2007-04-01 20:16:54 +04:00
|
|
|
#endif /* HAVE_SYS_WAIT_H */
|
2007-03-17 02:11:45 +03:00
|
|
|
#ifdef HAVE_STRING_H
|
|
|
|
#include <string.h>
|
|
|
|
#endif /* HAVE_STRING_H */
|
|
|
|
|
|
|
|
|
|
|
|
#include "opal/util/cmd_line.h"
|
2009-02-14 05:26:12 +03:00
|
|
|
#include "opal/util/output.h"
|
2007-03-17 02:11:45 +03:00
|
|
|
#include "opal/util/argv.h"
|
|
|
|
#include "opal/util/opal_environ.h"
|
|
|
|
#include "opal/mca/base/base.h"
|
|
|
|
#include "opal/mca/base/mca_base_param.h"
|
|
|
|
#include "opal/mca/crs/crs.h"
|
|
|
|
#include "opal/mca/crs/base/base.h"
|
|
|
|
#include "opal/runtime/opal.h"
|
|
|
|
#include "opal/runtime/opal_cr.h"
|
|
|
|
|
|
|
|
#include "orte/runtime/runtime.h"
|
|
|
|
#include "orte/runtime/orte_cr.h"
|
2008-02-28 04:57:57 +03:00
|
|
|
#include "orte/util/hnp_contact.h"
|
|
|
|
#include "orte/runtime/orte_globals.h"
|
|
|
|
#include "orte/util/name_fns.h"
|
2008-06-09 18:53:58 +04:00
|
|
|
#include "orte/util/show_help.h"
|
2007-03-17 02:11:45 +03:00
|
|
|
#include "orte/mca/rml/rml.h"
|
2009-02-14 05:26:12 +03:00
|
|
|
#include "orte/mca/rml/rml_types.h"
|
2008-02-28 04:57:57 +03:00
|
|
|
#include "orte/mca/errmgr/errmgr.h"
|
|
|
|
#include "opal/dss/dss.h"
|
2007-03-17 02:11:45 +03:00
|
|
|
#include "orte/mca/snapc/snapc.h"
|
|
|
|
#include "orte/mca/snapc/base/base.h"
|
|
|
|
|
2009-04-30 20:55:39 +04:00
|
|
|
#include MCA_timer_IMPLEMENTATION_HEADER
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
/******************
|
|
|
|
* Local Functions
|
|
|
|
******************/
|
|
|
|
static int ckpt_init(int argc, char *argv[]); /* Initalization routine */
|
|
|
|
static int ckpt_finalize(void); /* Finalization routine */
|
|
|
|
static int parse_args(int argc, char *argv[]);
|
2008-02-28 04:57:57 +03:00
|
|
|
static int find_hnp(void);
|
|
|
|
|
|
|
|
static int start_listener(void);
|
|
|
|
static int stop_listener(void);
|
|
|
|
static void hnp_receiver(int status,
|
|
|
|
orte_process_name_t* sender,
|
|
|
|
opal_buffer_t* buffer,
|
|
|
|
orte_rml_tag_t tag,
|
|
|
|
void* cbdata);
|
|
|
|
|
|
|
|
static void process_ckpt_update_cmd(orte_process_name_t* sender,
|
|
|
|
opal_buffer_t* buffer);
|
|
|
|
|
|
|
|
static int notify_process_for_checkpoint(int term);
|
|
|
|
static int pretty_print_status(void);
|
|
|
|
static int pretty_print_reference(void);
|
|
|
|
|
|
|
|
static orte_hnp_contact_t *orterun_hnp = NULL;
|
|
|
|
static char * global_snapshot_handle = NULL;
|
|
|
|
static int global_sequence_num = 0;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
/*****************************************
|
|
|
|
* Global Vars for Command line Arguments
|
|
|
|
*****************************************/
|
2008-04-23 04:17:12 +04:00
|
|
|
static bool listener_started = false;
|
|
|
|
|
2009-04-30 20:55:39 +04:00
|
|
|
static double timer_start = 0;
|
|
|
|
static double timer_last = 0;
|
|
|
|
static double get_time(void);
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
typedef struct {
|
|
|
|
bool help;
|
|
|
|
int pid;
|
|
|
|
bool term;
|
|
|
|
bool verbose;
|
2009-04-30 20:55:39 +04:00
|
|
|
int verbose_level;
|
2008-02-28 04:57:57 +03:00
|
|
|
orte_jobid_t req_hnp; /**< User Requested HNP */
|
2007-03-17 02:11:45 +03:00
|
|
|
bool nowait; /* Do not wait for checkpoint to complete before returning */
|
|
|
|
bool status; /* Display status messages while checkpoint is progressing */
|
|
|
|
int output;
|
2007-06-22 20:14:25 +04:00
|
|
|
int ckpt_status;
|
2007-03-17 02:11:45 +03:00
|
|
|
} orte_checkpoint_globals_t;
|
|
|
|
|
|
|
|
orte_checkpoint_globals_t orte_checkpoint_globals;
|
|
|
|
|
|
|
|
opal_cmd_line_init_t cmd_line_opts[] = {
|
|
|
|
{ NULL, NULL, NULL,
|
|
|
|
'h', NULL, "help",
|
|
|
|
0,
|
|
|
|
&orte_checkpoint_globals.help, OPAL_CMD_LINE_TYPE_BOOL,
|
|
|
|
"This help message" },
|
|
|
|
|
|
|
|
{ NULL, NULL, NULL,
|
|
|
|
'v', NULL, "verbose",
|
|
|
|
0,
|
|
|
|
&orte_checkpoint_globals.verbose, OPAL_CMD_LINE_TYPE_BOOL,
|
|
|
|
"Be Verbose" },
|
|
|
|
|
2009-04-30 20:55:39 +04:00
|
|
|
{ NULL, NULL, NULL,
|
|
|
|
'V', NULL, NULL,
|
|
|
|
1,
|
|
|
|
&orte_checkpoint_globals.verbose_level, OPAL_CMD_LINE_TYPE_INT,
|
|
|
|
"Set the verbosity level (For additional debugging information)" },
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
{ NULL, NULL, NULL,
|
|
|
|
'\0', NULL, "term",
|
|
|
|
0,
|
|
|
|
&orte_checkpoint_globals.term, OPAL_CMD_LINE_TYPE_BOOL,
|
|
|
|
"Terminate the application after checkpoint" },
|
|
|
|
|
|
|
|
{ NULL, NULL, NULL,
|
|
|
|
'w', NULL, "nowait",
|
|
|
|
0,
|
|
|
|
&orte_checkpoint_globals.nowait, OPAL_CMD_LINE_TYPE_BOOL,
|
|
|
|
"Do not wait for the application to finish checkpointing before returning" },
|
|
|
|
|
|
|
|
{ NULL, NULL, NULL,
|
|
|
|
's', NULL, "status",
|
|
|
|
0,
|
|
|
|
&orte_checkpoint_globals.status, OPAL_CMD_LINE_TYPE_BOOL,
|
|
|
|
"Display status messages describing the progression of the checkpoint" },
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
{ "hnp-jobid", NULL, NULL,
|
|
|
|
'\0', NULL, "hnp-jobid",
|
2007-03-17 02:11:45 +03:00
|
|
|
1,
|
2008-02-28 04:57:57 +03:00
|
|
|
&orte_checkpoint_globals.req_hnp, OPAL_CMD_LINE_TYPE_INT,
|
|
|
|
"This should be the jobid of the HNP whose applications you wish "
|
2007-03-17 02:11:45 +03:00
|
|
|
"to checkpoint." },
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
{ "hnp-pid", NULL, NULL,
|
|
|
|
'\0', NULL, "hnp-pid",
|
|
|
|
1,
|
|
|
|
&orte_checkpoint_globals.pid, OPAL_CMD_LINE_TYPE_INT,
|
|
|
|
"This should be the pid of the mpirun whose applications you wish "
|
|
|
|
"to checkpoint." },
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
/* End of list */
|
|
|
|
{ NULL, NULL, NULL, '\0', NULL, NULL, 0,
|
|
|
|
NULL, OPAL_CMD_LINE_TYPE_NULL,
|
|
|
|
NULL }
|
|
|
|
};
|
|
|
|
|
|
|
|
int
|
|
|
|
main(int argc, char *argv[])
|
|
|
|
{
|
|
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
|
|
|
|
/***************
|
|
|
|
* Initialize
|
|
|
|
***************/
|
|
|
|
if (ORTE_SUCCESS != (ret = ckpt_init(argc, argv))) {
|
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
|
|
|
|
/***************************
|
|
|
|
* Find the HNP that we want to connect to, if it exists
|
|
|
|
***************************/
|
|
|
|
if (ORTE_SUCCESS != (ret = find_hnp())) {
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output(0,
|
|
|
|
"HNP with PID %d Not found!",
|
|
|
|
orte_checkpoint_globals.pid);
|
2008-02-28 04:57:57 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
/*******************************
|
|
|
|
* Checkpoint the requested PID
|
|
|
|
*******************************/
|
|
|
|
if( orte_checkpoint_globals.verbose ) {
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2007-03-17 02:11:45 +03:00
|
|
|
"orte_checkpoint: Checkpointing...");
|
2008-02-28 04:57:57 +03:00
|
|
|
if (0 < orte_checkpoint_globals.pid) {
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2007-03-17 02:11:45 +03:00
|
|
|
"\t PID %d",
|
|
|
|
orte_checkpoint_globals.pid);
|
2008-02-28 04:57:57 +03:00
|
|
|
} else if (ORTE_JOBID_INVALID != orte_checkpoint_globals.req_hnp){
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2008-02-28 04:57:57 +03:00
|
|
|
"\t Mpirun (%s)",
|
2009-03-06 00:50:47 +03:00
|
|
|
ORTE_JOBID_PRINT(orte_checkpoint_globals.req_hnp));
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2008-02-28 04:57:57 +03:00
|
|
|
"\t Connected to Mpirun %s",
|
2009-03-06 00:50:47 +03:00
|
|
|
ORTE_NAME_PRINT(&orterun_hnp->name));
|
2008-02-28 04:57:57 +03:00
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
if(orte_checkpoint_globals.term) {
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2007-03-17 02:11:45 +03:00
|
|
|
"\t Terminating after checkpoint\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if(ORTE_SUCCESS != (ret = notify_process_for_checkpoint( orte_checkpoint_globals.term)) ) {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "ckpt_failure", true,
|
2007-06-21 01:05:19 +04:00
|
|
|
orte_checkpoint_globals.pid, ret);
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/*
|
|
|
|
* Wait for the checkpoint to complete
|
|
|
|
*/
|
|
|
|
if(!orte_checkpoint_globals.nowait) {
|
|
|
|
while( ORTE_SNAPC_CKPT_STATE_FINISHED != orte_checkpoint_globals.ckpt_status &&
|
|
|
|
ORTE_SNAPC_CKPT_STATE_ERROR != orte_checkpoint_globals.ckpt_status ) {
|
|
|
|
opal_progress();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-06-22 20:14:25 +04:00
|
|
|
if( ORTE_SNAPC_CKPT_STATE_ERROR == orte_checkpoint_globals.ckpt_status ) {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "ckpt_failure", true,
|
2007-06-22 20:14:25 +04:00
|
|
|
orte_checkpoint_globals.pid, ORTE_ERROR);
|
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2007-04-15 18:28:56 +04:00
|
|
|
if( orte_checkpoint_globals.status ) {
|
2008-02-28 04:57:57 +03:00
|
|
|
orte_checkpoint_globals.ckpt_status = ORTE_SNAPC_CKPT_STATE_FINISHED;
|
|
|
|
pretty_print_status();
|
2007-04-15 18:28:56 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
if(!orte_checkpoint_globals.nowait) {
|
2008-02-28 04:57:57 +03:00
|
|
|
pretty_print_reference();
|
2007-04-15 18:28:56 +04:00
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
cleanup:
|
|
|
|
/***************
|
|
|
|
* Cleanup
|
|
|
|
***************/
|
|
|
|
if (ORTE_SUCCESS != (ret = ckpt_finalize())) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return exit_status;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int parse_args(int argc, char *argv[]) {
|
|
|
|
int i, ret, len, exit_status = ORTE_SUCCESS ;
|
|
|
|
opal_cmd_line_t cmd_line;
|
|
|
|
char **app_env = NULL, **global_env = NULL;
|
2008-04-23 04:17:12 +04:00
|
|
|
char * tmp_env_var = NULL;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
/* Init structure */
|
|
|
|
memset(&orte_checkpoint_globals, 0, sizeof(orte_checkpoint_globals_t));
|
|
|
|
orte_checkpoint_globals.help = false;
|
|
|
|
orte_checkpoint_globals.pid = -1;
|
|
|
|
orte_checkpoint_globals.term = false;
|
|
|
|
orte_checkpoint_globals.verbose = false;
|
2009-04-30 20:55:39 +04:00
|
|
|
orte_checkpoint_globals.verbose_level = 0;
|
2008-02-28 04:57:57 +03:00
|
|
|
orte_checkpoint_globals.req_hnp = ORTE_JOBID_INVALID;
|
2007-03-17 02:11:45 +03:00
|
|
|
orte_checkpoint_globals.nowait = false;
|
|
|
|
orte_checkpoint_globals.status = false;
|
|
|
|
orte_checkpoint_globals.output = -1;
|
2007-06-22 20:14:25 +04:00
|
|
|
orte_checkpoint_globals.ckpt_status = ORTE_SNAPC_CKPT_STATE_NONE;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
/* Parse the command line options */
|
|
|
|
opal_cmd_line_create(&cmd_line, cmd_line_opts);
|
|
|
|
mca_base_open();
|
|
|
|
mca_base_cmd_line_setup(&cmd_line);
|
|
|
|
ret = opal_cmd_line_parse(&cmd_line, true, argc, argv);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Put all of the MCA arguments in the environment
|
|
|
|
*/
|
|
|
|
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
|
|
|
|
|
|
|
|
len = opal_argv_count(app_env);
|
|
|
|
for(i = 0; i < len; ++i) {
|
|
|
|
putenv(app_env[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
len = opal_argv_count(global_env);
|
|
|
|
for(i = 0; i < len; ++i) {
|
|
|
|
putenv(global_env[i]);
|
|
|
|
}
|
|
|
|
|
2008-04-23 04:17:12 +04:00
|
|
|
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
|
|
|
|
opal_setenv(tmp_env_var,
|
2007-03-17 02:11:45 +03:00
|
|
|
"1",
|
|
|
|
true, &environ);
|
2008-04-23 04:17:12 +04:00
|
|
|
free(tmp_env_var);
|
|
|
|
tmp_env_var = NULL;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
/**
|
|
|
|
* Now start parsing our specific arguments
|
|
|
|
*/
|
|
|
|
/* get the remaining bits */
|
|
|
|
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
|
|
|
|
|
2008-04-23 04:17:12 +04:00
|
|
|
#if OPAL_ENABLE_FT == 0
|
|
|
|
/* Warn and exit if not configured with Checkpoint/Restart */
|
|
|
|
{
|
|
|
|
char *args = NULL;
|
|
|
|
args = opal_cmd_line_get_usage_msg(&cmd_line);
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "usage-no-cr",
|
2008-04-23 04:17:12 +04:00
|
|
|
true, args);
|
|
|
|
free(args);
|
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
if (OPAL_SUCCESS != ret ||
|
|
|
|
orte_checkpoint_globals.help ||
|
2008-02-28 04:57:57 +03:00
|
|
|
(0 >= argc && ORTE_JOBID_INVALID == orte_checkpoint_globals.req_hnp)) {
|
2007-03-17 02:11:45 +03:00
|
|
|
char *args = NULL;
|
|
|
|
args = opal_cmd_line_get_usage_msg(&cmd_line);
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "usage", true,
|
2007-03-17 02:11:45 +03:00
|
|
|
args);
|
|
|
|
free(args);
|
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
2008-04-23 04:17:12 +04:00
|
|
|
|
2009-04-30 20:55:39 +04:00
|
|
|
if(orte_checkpoint_globals.verbose_level < 0 ) {
|
|
|
|
orte_checkpoint_globals.verbose_level = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(orte_checkpoint_globals.verbose_level > 0) {
|
|
|
|
orte_checkpoint_globals.verbose = true;
|
|
|
|
}
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
/*
|
2008-02-28 04:57:57 +03:00
|
|
|
* If the user did not supply an hnp jobid, then they must
|
2007-03-17 02:11:45 +03:00
|
|
|
* supply the PID of MPIRUN
|
|
|
|
*/
|
|
|
|
if(0 >= argc &&
|
2008-02-28 04:57:57 +03:00
|
|
|
ORTE_JOBID_INVALID != orte_checkpoint_globals.req_hnp) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ORTE_SUCCESS;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
|
|
|
orte_checkpoint_globals.pid = atoi(argv[0]);
|
|
|
|
if ( 0 >= orte_checkpoint_globals.pid ) {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "invalid_pid", true,
|
2007-03-17 02:11:45 +03:00
|
|
|
orte_checkpoint_globals.pid);
|
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* JJH: No wait is currently not implemented or tested
|
|
|
|
*/
|
|
|
|
if(orte_checkpoint_globals.nowait) {
|
|
|
|
orte_checkpoint_globals.nowait = false;
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "not_impl",
|
2007-03-17 02:11:45 +03:00
|
|
|
true,
|
|
|
|
"Disconnected checkpoint");
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if(orte_checkpoint_globals.verbose) {
|
|
|
|
orte_checkpoint_globals.status = true;
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
cleanup:
|
|
|
|
return exit_status;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2008-02-28 04:57:57 +03:00
|
|
|
* This function attempts to find an HNP to connect to.
|
2007-03-17 02:11:45 +03:00
|
|
|
*/
|
2008-02-28 04:57:57 +03:00
|
|
|
static int find_hnp(void) {
|
2008-06-04 18:44:11 +04:00
|
|
|
int ret, exit_status = ORTE_SUCCESS;
|
2008-02-28 04:57:57 +03:00
|
|
|
opal_list_t hnp_list;
|
|
|
|
opal_list_item_t *item;
|
|
|
|
orte_hnp_contact_t *hnpcandidate;
|
|
|
|
|
2008-12-10 20:10:39 +03:00
|
|
|
/* get the list of local hnp's available to us and setup
|
|
|
|
* contact info for them into the RML
|
|
|
|
*/
|
2008-02-28 04:57:57 +03:00
|
|
|
OBJ_CONSTRUCT(&hnp_list, opal_list_t);
|
2008-12-10 20:10:39 +03:00
|
|
|
if (ORTE_SUCCESS != (ret = orte_list_local_hnps(&hnp_list, true) ) ) {
|
2008-02-28 04:57:57 +03:00
|
|
|
ORTE_ERROR_LOG(ret);
|
2008-06-04 18:44:11 +04:00
|
|
|
exit_status = ret;
|
2007-03-17 02:11:45 +03:00
|
|
|
goto cleanup;
|
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
|
|
|
|
/* search the list for the desired hnp */
|
|
|
|
while (NULL != (item = opal_list_remove_first(&hnp_list))) {
|
|
|
|
hnpcandidate = (orte_hnp_contact_t*)item;
|
|
|
|
if (hnpcandidate->name.jobid == orte_checkpoint_globals.req_hnp ||
|
|
|
|
hnpcandidate->pid == orte_checkpoint_globals.pid) {
|
|
|
|
/* this is the one we want */
|
|
|
|
orterun_hnp = hnpcandidate;
|
2008-06-04 18:44:11 +04:00
|
|
|
exit_status = ORTE_SUCCESS;
|
2008-02-28 04:57:57 +03:00
|
|
|
goto cleanup;
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
cleanup:
|
|
|
|
while (NULL != (item = opal_list_remove_first(&hnp_list))) {
|
|
|
|
OBJ_RELEASE(item);
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
OBJ_DESTRUCT(&hnp_list);
|
2008-06-04 18:44:11 +04:00
|
|
|
|
|
|
|
if( NULL == orterun_hnp ) {
|
|
|
|
return ORTE_ERROR;
|
|
|
|
} else {
|
|
|
|
return exit_status;
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
static int ckpt_init(int argc, char *argv[]) {
|
|
|
|
int exit_status = ORTE_SUCCESS, ret;
|
2008-04-23 04:17:12 +04:00
|
|
|
char * tmp_env_var = NULL;
|
|
|
|
|
|
|
|
listener_started = false;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2007-04-22 02:38:15 +04:00
|
|
|
/*
|
|
|
|
* Make sure to init util before parse_args
|
|
|
|
* to ensure installdirs is setup properly
|
|
|
|
* before calling mca_base_open();
|
|
|
|
*/
|
|
|
|
if( ORTE_SUCCESS != (ret = opal_init_util()) ) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
/*
|
|
|
|
* Parse Command Line Arguments
|
|
|
|
*/
|
|
|
|
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Disable the checkpoint notification routine for this
|
|
|
|
* tool. As we will never need to checkpoint this tool.
|
|
|
|
* Note: This must happen before opal_init().
|
|
|
|
*/
|
|
|
|
opal_cr_set_enabled(false);
|
|
|
|
|
|
|
|
/* Select the none component, since we don't actually use a checkpointer */
|
2008-04-23 04:17:12 +04:00
|
|
|
tmp_env_var = mca_base_param_env_var("crs");
|
|
|
|
opal_setenv(tmp_env_var,
|
2007-03-17 02:11:45 +03:00
|
|
|
"none",
|
|
|
|
true, &environ);
|
2008-04-23 04:17:12 +04:00
|
|
|
free(tmp_env_var);
|
|
|
|
tmp_env_var = NULL;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
/***************************
|
2008-02-28 04:57:57 +03:00
|
|
|
* We need all of OPAL and the TOOLS portion of ORTE - this
|
|
|
|
* sets us up so we can talk to any HNP over the wire
|
2007-03-17 02:11:45 +03:00
|
|
|
***************************/
|
2008-02-28 04:57:57 +03:00
|
|
|
if (ORTE_SUCCESS != (ret = orte_init(ORTE_TOOL))) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-06-04 18:44:11 +04:00
|
|
|
/*
|
|
|
|
* Setup ORTE Output handle from the verbose argument
|
|
|
|
*/
|
|
|
|
if( orte_checkpoint_globals.verbose ) {
|
|
|
|
orte_checkpoint_globals.output = opal_output_open(NULL);
|
2009-04-30 20:55:39 +04:00
|
|
|
opal_output_set_verbosity(orte_checkpoint_globals.output, orte_checkpoint_globals.verbose_level);
|
2008-06-04 18:44:11 +04:00
|
|
|
} else {
|
|
|
|
orte_checkpoint_globals.output = 0; /* Default=STDERR */
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/*
|
|
|
|
* Start the listener
|
2007-03-17 02:11:45 +03:00
|
|
|
*/
|
2008-02-28 04:57:57 +03:00
|
|
|
if( ORTE_SUCCESS != (ret = start_listener() ) ) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
cleanup:
|
|
|
|
return exit_status;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ckpt_finalize(void) {
|
|
|
|
int exit_status = ORTE_SUCCESS, ret;
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/*
|
|
|
|
* Stop the listener
|
|
|
|
*/
|
|
|
|
if( ORTE_SUCCESS != (ret = stop_listener() ) ) {
|
|
|
|
exit_status = ret;
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
|
|
|
|
if (ORTE_SUCCESS != (ret = orte_finalize())) {
|
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
2007-05-25 01:27:06 +04:00
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
cleanup:
|
|
|
|
return exit_status;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
static int start_listener(void)
|
|
|
|
{
|
2007-03-17 02:11:45 +03:00
|
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if (ORTE_SUCCESS != (ret = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
|
|
|
|
ORTE_RML_TAG_CKPT,
|
|
|
|
ORTE_RML_PERSISTENT,
|
|
|
|
hnp_receiver,
|
|
|
|
NULL))) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-04-23 04:17:12 +04:00
|
|
|
listener_started = true;
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
cleanup:
|
|
|
|
return exit_status;
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
static int stop_listener(void)
|
|
|
|
{
|
|
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
|
2008-04-23 04:17:12 +04:00
|
|
|
if( !listener_started ) {
|
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if (ORTE_SUCCESS != (ret = orte_rml.recv_cancel(ORTE_NAME_WILDCARD,
|
|
|
|
ORTE_RML_TAG_CKPT))) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-04-23 04:17:12 +04:00
|
|
|
listener_started = false;
|
2008-02-28 04:57:57 +03:00
|
|
|
cleanup:
|
|
|
|
return exit_status;
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
static void hnp_receiver(int status,
|
|
|
|
orte_process_name_t* sender,
|
|
|
|
opal_buffer_t* buffer,
|
|
|
|
orte_rml_tag_t tag,
|
|
|
|
void* cbdata)
|
|
|
|
{
|
|
|
|
orte_snapc_cmd_flag_t command;
|
|
|
|
orte_std_cntr_t count;
|
|
|
|
int rc;
|
|
|
|
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(5, orte_checkpoint_globals.output,
|
2008-02-28 04:57:57 +03:00
|
|
|
"orte_checkpoint: hnp_receiver: Receive a command message.");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Otherwise this is an inter-coordinator command (usually updating state info).
|
|
|
|
*/
|
|
|
|
count = 1;
|
|
|
|
if (ORTE_SUCCESS != (rc = opal_dss.unpack(buffer, &command, &count, ORTE_SNAPC_CMD))) {
|
|
|
|
ORTE_ERROR_LOG(rc);
|
|
|
|
return;
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
|
|
|
|
switch (command) {
|
|
|
|
case ORTE_SNAPC_GLOBAL_UPDATE_CMD:
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2008-02-28 04:57:57 +03:00
|
|
|
"orte_checkpoint: hnp_receiver: Status Update.");
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
process_ckpt_update_cmd(sender, buffer);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case ORTE_SNAPC_GLOBAL_INIT_CMD:
|
|
|
|
case ORTE_SNAPC_GLOBAL_TERM_CMD:
|
|
|
|
/* Do Nothing */
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
ORTE_ERROR_LOG(ORTE_ERR_VALUE_OUT_OF_BOUNDS);
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
static void process_ckpt_update_cmd(orte_process_name_t* sender,
|
|
|
|
opal_buffer_t* buffer)
|
|
|
|
{
|
|
|
|
int ret, exit_status = ORTE_SUCCESS;
|
|
|
|
orte_std_cntr_t count = 1;
|
|
|
|
int ckpt_status = ORTE_SNAPC_CKPT_STATE_NONE;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/*
|
|
|
|
* Receive the data:
|
|
|
|
* - ckpt_state
|
|
|
|
* - global snapshot handle (upon finish only)
|
|
|
|
* - sequence number (upon finish only)
|
|
|
|
*/
|
|
|
|
count = 1;
|
|
|
|
if ( ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &ckpt_status, &count, OPAL_INT)) ) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
orte_checkpoint_globals.ckpt_status = ckpt_status;
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if( ORTE_SNAPC_CKPT_STATE_FINISHED == orte_checkpoint_globals.ckpt_status ||
|
|
|
|
ORTE_SNAPC_CKPT_STATE_ERROR == orte_checkpoint_globals.ckpt_status ) {
|
|
|
|
count = 1;
|
|
|
|
if ( ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &global_snapshot_handle, &count, OPAL_STRING)) ) {
|
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
count = 1;
|
|
|
|
if ( ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &global_sequence_num, &count, OPAL_INT)) ) {
|
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/*
|
|
|
|
* If the job is not able to be checkpointed, then return
|
|
|
|
*/
|
|
|
|
if( ORTE_SNAPC_CKPT_STATE_NO_CKPT == orte_checkpoint_globals.ckpt_status) {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "non-ckptable",
|
2008-02-28 04:57:57 +03:00
|
|
|
true,
|
|
|
|
orte_checkpoint_globals.pid);
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
2008-02-28 04:57:57 +03:00
|
|
|
/*
|
|
|
|
* If we are to display the status progression
|
|
|
|
*/
|
|
|
|
if( orte_checkpoint_globals.status ) {
|
|
|
|
if(ORTE_SNAPC_CKPT_STATE_FINISHED != orte_checkpoint_globals.ckpt_status) {
|
|
|
|
pretty_print_status();
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
cleanup:
|
2008-02-28 04:57:57 +03:00
|
|
|
return;
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
static int
|
|
|
|
notify_process_for_checkpoint(int term)
|
|
|
|
{
|
2007-03-17 02:11:45 +03:00
|
|
|
int ret, exit_status = ORTE_SUCCESS;
|
2008-02-28 04:57:57 +03:00
|
|
|
opal_buffer_t *buffer = NULL;
|
|
|
|
orte_snapc_cmd_flag_t command = ORTE_SNAPC_GLOBAL_INIT_CMD;
|
|
|
|
orte_jobid_t jobid = ORTE_JOBID_INVALID;
|
2007-06-22 20:14:25 +04:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if (NULL == (buffer = OBJ_NEW(opal_buffer_t))) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ORTE_ERROR;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2008-02-28 04:57:57 +03:00
|
|
|
"orte_checkpoint: notify_hnp: Contact Head Node Process PID %d\n",
|
|
|
|
orte_checkpoint_globals.pid);
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2009-04-30 20:55:39 +04:00
|
|
|
timer_start = get_time();
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/***********************************
|
|
|
|
* Notify HNP of checkpoint request
|
|
|
|
* Send:
|
|
|
|
* - Command
|
|
|
|
* - term flag
|
|
|
|
* - jobid
|
|
|
|
***********************************/
|
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &command, 1, ORTE_SNAPC_CMD)) ) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &term, 1, OPAL_BOOL))) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &jobid, 1, ORTE_JOBID))) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if ( 0 > (ret = orte_rml.send_buffer(&(orterun_hnp->name), buffer, ORTE_RML_TAG_CKPT, 0)) ) {
|
2007-03-17 02:11:45 +03:00
|
|
|
exit_status = ret;
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
2008-06-04 18:44:11 +04:00
|
|
|
opal_output_verbose(10, orte_checkpoint_globals.output,
|
2008-02-28 04:57:57 +03:00
|
|
|
"orte_checkpoint: notify_hnp: Requested a checkpoint of jobid %s\n",
|
2009-03-06 00:50:47 +03:00
|
|
|
ORTE_JOBID_PRINT(jobid));
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
cleanup:
|
|
|
|
if( NULL != buffer) {
|
|
|
|
OBJ_RELEASE(buffer);
|
|
|
|
buffer = NULL;
|
2007-03-17 02:11:45 +03:00
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
if( ORTE_SUCCESS != exit_status ) {
|
This commit represents a bunch of work on a Mercurial side branch. As
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
2008-05-14 00:00:55 +04:00
|
|
|
orte_show_help("help-orte-checkpoint.txt", "unable_to_connect", true,
|
2008-02-28 04:57:57 +03:00
|
|
|
orte_checkpoint_globals.pid);
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
return exit_status;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
/***************
|
|
|
|
* Pretty Print
|
|
|
|
***************/
|
2009-04-30 20:55:39 +04:00
|
|
|
static double 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;
|
|
|
|
}
|
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
static int pretty_print_status(void) {
|
2007-03-17 02:11:45 +03:00
|
|
|
char * state_str = NULL;
|
2009-04-30 20:55:39 +04:00
|
|
|
double cur_time;
|
|
|
|
|
|
|
|
cur_time = get_time();
|
|
|
|
|
|
|
|
if( timer_last == 0 ) {
|
|
|
|
timer_last = cur_time;
|
|
|
|
}
|
|
|
|
|
|
|
|
orte_snapc_ckpt_state_str(&state_str, orte_checkpoint_globals.ckpt_status);
|
2007-03-17 02:11:45 +03:00
|
|
|
|
2009-04-30 20:55:39 +04:00
|
|
|
if( NULL != global_snapshot_handle ) {
|
|
|
|
opal_output(0,
|
|
|
|
"[%6.2f / %6.2f] %*s - %s\n",
|
|
|
|
(cur_time - timer_last), (cur_time - timer_start),
|
|
|
|
25, state_str, global_snapshot_handle);
|
|
|
|
} else {
|
|
|
|
opal_output(0,
|
|
|
|
"[%6.2f / %6.2f] %*s - ...\n",
|
|
|
|
(cur_time - timer_last), (cur_time - timer_start),
|
|
|
|
25, state_str);
|
|
|
|
}
|
2007-03-17 02:11:45 +03:00
|
|
|
|
|
|
|
if( NULL != state_str) {
|
|
|
|
free(state_str);
|
|
|
|
}
|
2009-04-30 20:55:39 +04:00
|
|
|
|
|
|
|
timer_last = cur_time;
|
|
|
|
|
2007-03-17 02:11:45 +03:00
|
|
|
return ORTE_SUCCESS;
|
|
|
|
}
|
2007-04-15 18:28:56 +04:00
|
|
|
|
2008-02-28 04:57:57 +03:00
|
|
|
static int pretty_print_reference(void) {
|
2007-04-15 18:28:56 +04:00
|
|
|
|
2007-04-25 23:54:53 +04:00
|
|
|
printf("Snapshot Ref.: %3d %s\n",
|
2008-02-28 04:57:57 +03:00
|
|
|
global_sequence_num,
|
|
|
|
global_snapshot_handle);
|
2007-04-15 18:28:56 +04:00
|
|
|
|
|
|
|
return ORTE_SUCCESS;
|
|
|
|
}
|