Add a new ESS module for use by local slave processes - only active when specifically selected
This commit was SVN r20438.
Этот коммит содержится в:
Ralph Castain
родитель
b7e6bafada
Коммит
6292b797e9
43
orte/mca/ess/slave/Makefile.am
Обычный файл
43
orte/mca/ess/slave/Makefile.am
Обычный файл
@ -0,0 +1,43 @@
|
||||
#
|
||||
# Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
# 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.
|
||||
# $COPYRIGHT$
|
||||
#
|
||||
# Additional copyrights may follow
|
||||
#
|
||||
# $HEADER$
|
||||
#
|
||||
|
||||
sources = \
|
||||
ess_slave.h \
|
||||
ess_slave_component.c \
|
||||
ess_slave_module.c
|
||||
|
||||
# Make the output library in this directory, and name it either
|
||||
# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
|
||||
# (for static builds).
|
||||
|
||||
if OMPI_BUILD_ess_slave_DSO
|
||||
component_noinst =
|
||||
component_install = mca_ess_slave.la
|
||||
else
|
||||
component_noinst = libmca_ess_slave.la
|
||||
component_install =
|
||||
endif
|
||||
|
||||
mcacomponentdir = $(pkglibdir)
|
||||
mcacomponent_LTLIBRARIES = $(component_install)
|
||||
mca_ess_slave_la_SOURCES = $(sources)
|
||||
mca_ess_slave_la_LDFLAGS = -module -avoid-version
|
||||
|
||||
noinst_LTLIBRARIES = $(component_noinst)
|
||||
libmca_ess_slave_la_SOURCES =$(sources)
|
||||
libmca_ess_slave_la_LDFLAGS = -module -avoid-version
|
||||
14
orte/mca/ess/slave/configure.m4
Обычный файл
14
orte/mca/ess/slave/configure.m4
Обычный файл
@ -0,0 +1,14 @@
|
||||
# -*- shell-script -*-
|
||||
#
|
||||
# Copyright (c) 2009 Los Alamos National Security, LLC.
|
||||
# All rights reserved.
|
||||
# $COPYRIGHT$
|
||||
#
|
||||
# Additional copyrights may follow
|
||||
#
|
||||
# $HEADER$
|
||||
#
|
||||
|
||||
# MCA_ess_slave_CONFIG([action-if-found], [action-if-not-found])
|
||||
# -----------------------------------------------------------
|
||||
AC_DEFUN([MCA_ess_slave_CONFIG], [$1])
|
||||
27
orte/mca/ess/slave/configure.params
Обычный файл
27
orte/mca/ess/slave/configure.params
Обычный файл
@ -0,0 +1,27 @@
|
||||
# -*- shell-script -*-
|
||||
#
|
||||
# Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
# 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.
|
||||
# Copyright (c) 2009 Los Alamos National Security, LLC. All rights
|
||||
# reserved.
|
||||
# $COPYRIGHT$
|
||||
#
|
||||
# Additional copyrights may follow
|
||||
#
|
||||
# $HEADER$
|
||||
#
|
||||
|
||||
PARAM_CONFIG_FILES="Makefile"
|
||||
#
|
||||
# Set the config priority so that we build
|
||||
# whenever someone using daemons does
|
||||
|
||||
PARAM_CONFIG_PRIORITY=10
|
||||
36
orte/mca/ess/slave/ess_slave.h
Обычный файл
36
orte/mca/ess/slave/ess_slave.h
Обычный файл
@ -0,0 +1,36 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
|
||||
* University Research and Technology
|
||||
* Corporation. All rights reserved.
|
||||
* Copyright (c) 2004-2006 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.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*/
|
||||
|
||||
#ifndef ORTE_ESS_SLAVE_H
|
||||
#define ORTE_ESS_SLAVE_H
|
||||
|
||||
BEGIN_C_DECLS
|
||||
|
||||
/*
|
||||
* Module open / close
|
||||
*/
|
||||
int orte_ess_slave_component_open(void);
|
||||
int orte_ess_slave_component_close(void);
|
||||
int orte_ess_slave_component_query(mca_base_module_t **module, int *priority);
|
||||
|
||||
|
||||
ORTE_MODULE_DECLSPEC extern orte_ess_base_component_t mca_ess_slave_component;
|
||||
|
||||
END_C_DECLS
|
||||
|
||||
#endif /* ORTE_ESS_SLAVE_H */
|
||||
86
orte/mca/ess/slave/ess_slave_component.c
Обычный файл
86
orte/mca/ess/slave/ess_slave_component.c
Обычный файл
@ -0,0 +1,86 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
|
||||
* 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.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*
|
||||
* These symbols are in a file by themselves to provide nice linker
|
||||
* semantics. Since linkers generally pull in symbols by object
|
||||
* files, keeping these symbols as the only symbols in this file
|
||||
* prevents utility programs such as "ompi_info" from having to import
|
||||
* entire components just to query their version and parameters.
|
||||
*/
|
||||
|
||||
#include "orte_config.h"
|
||||
#include "orte/constants.h"
|
||||
|
||||
#include "opal/mca/base/mca_base_param.h"
|
||||
|
||||
#include "orte/util/proc_info.h"
|
||||
|
||||
#include "orte/mca/ess/ess.h"
|
||||
#include "orte/mca/ess/slave/ess_slave.h"
|
||||
|
||||
extern orte_ess_base_module_t orte_ess_slave_module;
|
||||
|
||||
/*
|
||||
* Instantiate the public struct with all of our public information
|
||||
* and pointers to our public functions in it
|
||||
*/
|
||||
orte_ess_base_component_t mca_ess_slave_component = {
|
||||
{
|
||||
ORTE_ESS_BASE_VERSION_2_0_0,
|
||||
|
||||
/* Component name and version */
|
||||
"slave",
|
||||
ORTE_MAJOR_VERSION,
|
||||
ORTE_MINOR_VERSION,
|
||||
ORTE_RELEASE_VERSION,
|
||||
|
||||
/* Component open and close functions */
|
||||
orte_ess_slave_component_open,
|
||||
orte_ess_slave_component_close,
|
||||
orte_ess_slave_component_query
|
||||
},
|
||||
{
|
||||
/* The component is checkpoint ready */
|
||||
MCA_BASE_METADATA_PARAM_CHECKPOINT
|
||||
}
|
||||
};
|
||||
|
||||
|
||||
int
|
||||
orte_ess_slave_component_open(void)
|
||||
{
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
int orte_ess_slave_component_query(mca_base_module_t **module, int *priority)
|
||||
{
|
||||
/* we are the slave module, so set the priority so
|
||||
* we can only be selected if directed to do so
|
||||
*/
|
||||
|
||||
*priority = 0;
|
||||
*module = (mca_base_module_t *)&orte_ess_slave_module;
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
|
||||
int
|
||||
orte_ess_slave_component_close(void)
|
||||
{
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
544
orte/mca/ess/slave/ess_slave_module.c
Обычный файл
544
orte/mca/ess/slave/ess_slave_module.c
Обычный файл
@ -0,0 +1,544 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
* 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.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*
|
||||
*/
|
||||
|
||||
#include "orte_config.h"
|
||||
#include "orte/constants.h"
|
||||
|
||||
#include <sys/types.h>
|
||||
#include <stdio.h>
|
||||
#ifdef HAVE_FCNTL_H
|
||||
#include <fcntl.h>
|
||||
#endif
|
||||
#ifdef HAVE_UNISTD_H
|
||||
#include <unistd.h>
|
||||
#endif
|
||||
|
||||
#include "opal/event/event.h"
|
||||
#include "opal/threads/mutex.h"
|
||||
#include "opal/runtime/opal.h"
|
||||
#include "opal/runtime/opal_cr.h"
|
||||
#include "opal/class/opal_pointer_array.h"
|
||||
|
||||
#include "orte/util/show_help.h"
|
||||
#include "opal/mca/mca.h"
|
||||
#include "opal/mca/base/base.h"
|
||||
#include "opal/mca/base/mca_base_param.h"
|
||||
#include "opal/util/os_path.h"
|
||||
#include "opal/util/cmd_line.h"
|
||||
#include "opal/util/malloc.h"
|
||||
#include "opal/util/argv.h"
|
||||
|
||||
#include "orte/mca/rml/base/base.h"
|
||||
#include "orte/mca/rml/base/rml_contact.h"
|
||||
#include "orte/mca/routed/base/base.h"
|
||||
#include "orte/mca/routed/routed.h"
|
||||
#include "orte/mca/errmgr/base/base.h"
|
||||
#include "orte/mca/grpcomm/base/base.h"
|
||||
#include "orte/mca/iof/base/base.h"
|
||||
#include "orte/mca/ess/base/base.h"
|
||||
#include "orte/mca/ess/ess.h"
|
||||
#include "orte/mca/ras/base/base.h"
|
||||
#include "orte/mca/plm/base/base.h"
|
||||
#include "orte/mca/odls/base/base.h"
|
||||
|
||||
#include "orte/mca/rmaps/base/base.h"
|
||||
#if OPAL_ENABLE_FT == 1
|
||||
#include "orte/mca/snapc/base/base.h"
|
||||
#endif
|
||||
#include "orte/mca/filem/base/base.h"
|
||||
#include "orte/util/proc_info.h"
|
||||
#include "orte/util/session_dir.h"
|
||||
#include "orte/util/hnp_contact.h"
|
||||
#include "orte/util/name_fns.h"
|
||||
#include "orte/util/nidmap.h"
|
||||
|
||||
#include "orte/runtime/runtime.h"
|
||||
#include "orte/runtime/orte_wait.h"
|
||||
#include "orte/runtime/orte_globals.h"
|
||||
|
||||
#include "orte/runtime/orte_cr.h"
|
||||
#include "orte/mca/ess/ess.h"
|
||||
#include "orte/mca/ess/base/base.h"
|
||||
#include "orte/mca/ess/slave/ess_slave.h"
|
||||
|
||||
static int slave_set_name(void);
|
||||
|
||||
static int rte_init(char flags);
|
||||
static int rte_finalize(void);
|
||||
static bool proc_is_local(orte_process_name_t *proc);
|
||||
static orte_vpid_t proc_get_daemon(orte_process_name_t *proc);
|
||||
static char* proc_get_hostname(orte_process_name_t *proc);
|
||||
static uint32_t proc_get_arch(orte_process_name_t *proc);
|
||||
static orte_local_rank_t proc_get_local_rank(orte_process_name_t *proc);
|
||||
static orte_node_rank_t proc_get_node_rank(orte_process_name_t *proc);
|
||||
static int update_arch(orte_process_name_t *proc, uint32_t arch);
|
||||
static int update_pidmap(opal_byte_object_t *bo);
|
||||
static int update_nidmap(opal_byte_object_t *bo);
|
||||
|
||||
#if OPAL_ENABLE_FT == 1
|
||||
static int rte_ft_event(int state);
|
||||
static int ess_slave_ft_event_update_process_info(orte_process_name_t proc, pid_t pid);
|
||||
#endif
|
||||
|
||||
orte_ess_base_module_t orte_ess_slave_module = {
|
||||
rte_init,
|
||||
rte_finalize,
|
||||
orte_ess_base_app_abort,
|
||||
proc_is_local,
|
||||
proc_get_daemon,
|
||||
proc_get_hostname,
|
||||
proc_get_arch,
|
||||
proc_get_local_rank,
|
||||
proc_get_node_rank,
|
||||
update_arch,
|
||||
update_pidmap,
|
||||
update_nidmap,
|
||||
#if OPAL_ENABLE_FT == 1
|
||||
rte_ft_event
|
||||
#else
|
||||
NULL
|
||||
#endif
|
||||
};
|
||||
|
||||
static int rte_init(char flags)
|
||||
{
|
||||
int ret;
|
||||
char *error = NULL;
|
||||
|
||||
/* run the prolog */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
|
||||
error = "orte_ess_base_std_prolog";
|
||||
goto error;
|
||||
}
|
||||
|
||||
/* Start by getting a unique name from the enviro */
|
||||
slave_set_name();
|
||||
|
||||
/* use the default procedure to finish my setup */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_app_setup())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
error = "orte_ess_base_app_setup";
|
||||
goto error;
|
||||
}
|
||||
|
||||
/* init my nidmap arrays - no data can be available, but
|
||||
* we want to ensure that nobody else who looks at
|
||||
* those arrays will segfault
|
||||
*/
|
||||
if (ORTE_SUCCESS != (ret = orte_util_nidmap_init(NULL))) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
error = "orte_util_nidmap_init";
|
||||
goto error;
|
||||
}
|
||||
|
||||
return ORTE_SUCCESS;
|
||||
|
||||
error:
|
||||
orte_show_help("help-orte-runtime.txt",
|
||||
"orte_init:startup:internal-failure",
|
||||
true, error, ORTE_ERROR_NAME(ret), ret);
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
static int rte_finalize(void)
|
||||
{
|
||||
int ret;
|
||||
|
||||
/* use the default procedure to finish */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_app_finalize())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
}
|
||||
|
||||
/* deconstruct the nidmap and jobmap arrays */
|
||||
orte_util_nidmap_finalize();
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
static bool proc_is_local(orte_process_name_t *proc)
|
||||
{
|
||||
/* no proc can be local */
|
||||
|
||||
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
|
||||
"%s ess:slave: proc %s is REMOTE",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
|
||||
ORTE_NAME_PRINT(proc)));
|
||||
|
||||
return false;
|
||||
|
||||
}
|
||||
|
||||
static orte_vpid_t proc_get_daemon(orte_process_name_t *proc)
|
||||
{
|
||||
/* if it is me, the answer is my daemon's vpid */
|
||||
if (proc->jobid == ORTE_PROC_MY_NAME->jobid &&
|
||||
proc->vpid == ORTE_PROC_MY_NAME->vpid) {
|
||||
return ORTE_PROC_MY_DAEMON->vpid;
|
||||
}
|
||||
|
||||
/* otherwise, no idea */
|
||||
return ORTE_VPID_INVALID;
|
||||
}
|
||||
|
||||
static char* proc_get_hostname(orte_process_name_t *proc)
|
||||
{
|
||||
/* if it is me, the answer is my nodename */
|
||||
if (proc->jobid == ORTE_PROC_MY_NAME->jobid &&
|
||||
proc->vpid == ORTE_PROC_MY_NAME->vpid) {
|
||||
return orte_process_info.nodename;
|
||||
}
|
||||
|
||||
/* otherwise, no idea */
|
||||
return NULL;
|
||||
}
|
||||
|
||||
static uint32_t proc_get_arch(orte_process_name_t *proc)
|
||||
{
|
||||
/* if it is me, the answer is my arch */
|
||||
if (proc->jobid == ORTE_PROC_MY_NAME->jobid &&
|
||||
proc->vpid == ORTE_PROC_MY_NAME->vpid) {
|
||||
return orte_process_info.arch;
|
||||
}
|
||||
|
||||
/* otherwise, no idea */
|
||||
return 0;
|
||||
}
|
||||
|
||||
static int update_arch(orte_process_name_t *proc, uint32_t arch)
|
||||
{
|
||||
/* nothing to do */
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
static orte_local_rank_t proc_get_local_rank(orte_process_name_t *proc)
|
||||
{
|
||||
/* if it is me, the local rank is zero */
|
||||
if (proc->jobid == ORTE_PROC_MY_NAME->jobid &&
|
||||
proc->vpid == ORTE_PROC_MY_NAME->vpid) {
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* otherwise, no idea */
|
||||
return ORTE_LOCAL_RANK_INVALID;
|
||||
}
|
||||
|
||||
static orte_node_rank_t proc_get_node_rank(orte_process_name_t *proc)
|
||||
{
|
||||
/* if it is me, the node rank is zero */
|
||||
if (proc->jobid == ORTE_PROC_MY_NAME->jobid &&
|
||||
proc->vpid == ORTE_PROC_MY_NAME->vpid) {
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* otherwise, no idea */
|
||||
return ORTE_NODE_RANK_INVALID;
|
||||
}
|
||||
|
||||
static int update_pidmap(opal_byte_object_t *bo)
|
||||
{
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
static int update_nidmap(opal_byte_object_t *bo)
|
||||
{
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
static int slave_set_name(void)
|
||||
{
|
||||
char *jobid_str, *procid_str;
|
||||
int id, rc;
|
||||
orte_jobid_t jobid;
|
||||
orte_vpid_t vpid;
|
||||
|
||||
id = mca_base_param_register_string("orte", "ess", "jobid", NULL, NULL);
|
||||
mca_base_param_lookup_string(id, &jobid_str);
|
||||
if (NULL == jobid_str) {
|
||||
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
|
||||
return ORTE_ERR_NOT_FOUND;
|
||||
}
|
||||
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, jobid_str))) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return(rc);
|
||||
}
|
||||
free(jobid_str);
|
||||
|
||||
id = mca_base_param_register_string("orte", "ess", "vpid", NULL, NULL);
|
||||
mca_base_param_lookup_string(id, &procid_str);
|
||||
if (NULL == procid_str) {
|
||||
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
|
||||
return ORTE_ERR_NOT_FOUND;
|
||||
}
|
||||
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, procid_str))) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return(rc);
|
||||
}
|
||||
free(procid_str);
|
||||
|
||||
ORTE_PROC_MY_NAME->jobid = jobid;
|
||||
ORTE_PROC_MY_NAME->vpid = vpid;
|
||||
|
||||
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
|
||||
"ess:slave set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
|
||||
|
||||
/* get the non-name common environmental variables */
|
||||
if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return rc;
|
||||
}
|
||||
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
#if OPAL_ENABLE_FT == 1
|
||||
static int rte_ft_event(int state)
|
||||
{
|
||||
int ret, exit_status = ORTE_SUCCESS;
|
||||
|
||||
/******** Checkpoint Prep ********/
|
||||
if(OPAL_CRS_CHECKPOINT == state) {
|
||||
/*
|
||||
* Notify SnapC
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_snapc.ft_event(OPAL_CRS_CHECKPOINT))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Notify Routed
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_routed.ft_event(OPAL_CRS_CHECKPOINT))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Notify RML -> OOB
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_rml.ft_event(OPAL_CRS_CHECKPOINT))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
}
|
||||
/******** Continue Recovery ********/
|
||||
else if (OPAL_CRS_CONTINUE == state ) {
|
||||
/*
|
||||
* Notify RML -> OOB
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_rml.ft_event(OPAL_CRS_CONTINUE))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Notify Routed
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_routed.ft_event(OPAL_CRS_CONTINUE))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Notify SnapC
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_snapc.ft_event(OPAL_CRS_CONTINUE))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
}
|
||||
/******** Restart Recovery ********/
|
||||
else if (OPAL_CRS_RESTART == state ) {
|
||||
/*
|
||||
* This should follow the ess init() function
|
||||
*/
|
||||
|
||||
/*
|
||||
* Clear nidmap and jmap
|
||||
*/
|
||||
orte_util_nidmap_finalize();
|
||||
|
||||
/*
|
||||
* - Reset Contact information
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = slave_set_name() ) ) {
|
||||
exit_status = ret;
|
||||
}
|
||||
|
||||
/*
|
||||
* Notify RML -> OOB
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_rml.ft_event(OPAL_CRS_RESTART))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Restart the routed framework
|
||||
* JJH: Lie to the finalize function so it does not try to contact the daemon.
|
||||
*/
|
||||
orte_process_info.tool = true;
|
||||
if (ORTE_SUCCESS != (ret = orte_routed.finalize()) ) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
orte_process_info.tool = false;
|
||||
if (ORTE_SUCCESS != (ret = orte_routed.initialize()) ) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Group Comm - Clean out stale data
|
||||
*/
|
||||
orte_grpcomm.finalize();
|
||||
if (ORTE_SUCCESS != (ret = orte_grpcomm.init())) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
if (ORTE_SUCCESS != (ret = orte_grpcomm.purge_proc_attrs())) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Restart the PLM - Does nothing at the moment, but included for completeness
|
||||
*/
|
||||
if (ORTE_SUCCESS != (ret = orte_plm.finalize())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
if (ORTE_SUCCESS != (ret = orte_plm.init())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* RML - Enable communications
|
||||
*/
|
||||
if (ORTE_SUCCESS != (ret = orte_rml.enable_comm())) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Session directory re-init
|
||||
*/
|
||||
if (ORTE_SUCCESS != (ret = orte_session_dir(true,
|
||||
orte_process_info.tmpdir_base,
|
||||
orte_process_info.nodename,
|
||||
NULL, /* Batch ID -- Not used */
|
||||
ORTE_PROC_MY_NAME))) {
|
||||
exit_status = ret;
|
||||
}
|
||||
|
||||
opal_output_set_output_file_info(orte_process_info.proc_session_dir,
|
||||
"output-", NULL, NULL);
|
||||
|
||||
/*
|
||||
* Notify Routed
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_routed.ft_event(OPAL_CRS_RESTART))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* Notify SnapC
|
||||
*/
|
||||
if( ORTE_SUCCESS != (ret = orte_snapc.ft_event(OPAL_CRS_RESTART))) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/*
|
||||
* 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 = ess_slave_ft_event_update_process_info(orte_process_info.my_name, getpid())) ) {
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
/* if one was provided, build my nidmap */
|
||||
if (ORTE_SUCCESS != (ret = orte_util_nidmap_init(orte_process_info.sync_buf))) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
exit_status = ret;
|
||||
goto cleanup;
|
||||
}
|
||||
}
|
||||
else if (OPAL_CRS_TERM == state ) {
|
||||
/* Nothing */
|
||||
}
|
||||
else {
|
||||
/* Error state = Nothing */
|
||||
}
|
||||
|
||||
cleanup:
|
||||
|
||||
return exit_status;
|
||||
}
|
||||
|
||||
static int ess_slave_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;
|
||||
}
|
||||
|
||||
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 (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;
|
||||
}
|
||||
#endif
|
||||
|
||||
Загрузка…
x
Ссылка в новой задаче
Block a user