1
1
openmpi/orte/mca/ess/env/ess_env_module.c
Ralph Castain c199dbb241 Revert r21480 - we already did open/select the PLM on non-HNP daemons. This commit broke slave launches on Torque and SLURM as it caused the PLM to be open/selected twice.
The open/select of the PLM is done in orte/mca/ess/base/ess_base_std_orted.c. It only is done when the PLM MCA param is set directing a specific PLM be selected. The function

orte_plm_base_orted_append_basic_args

clears the params passed to the daemon of any PLM selection passed to the HNP. Each PLM then adds a PLM directive if-and-only-if backend PLM support is desired. At present, Torque, SLURM, and rsh all specify this support and direct that the backend orted open the "rsh" PLM.

This commit was SVN r21488.

The following SVN revision numbers were found above:
  r21480 --> open-mpi/ompi@ed585bce8a
2009-06-20 03:58:00 +00:00

688 строки
19 KiB
C

/*
* 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 <stdlib.h>
#include "opal/event/event.h"
#include "opal/runtime/opal.h"
#include "opal/mca/paffinity/paffinity.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/output.h"
#include "opal/util/malloc.h"
#include "opal/util/argv.h"
#include "orte/mca/rml/base/base.h"
#include "orte/mca/rml/rml_types.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/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/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/env/ess_env.h"
static int env_set_name(void);
static int rte_init(void);
static int rte_finalize(void);
static uint8_t proc_get_locality(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_env_ft_event_update_process_info(orte_process_name_t proc, pid_t pid);
#endif
orte_ess_base_module_t orte_ess_env_module = {
rte_init,
rte_finalize,
orte_ess_base_app_abort,
proc_get_locality,
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
};
/*
* Local variables
*/
static orte_node_rank_t my_node_rank=ORTE_NODE_RANK_INVALID;
static int rte_init(void)
{
int ret;
char *error = NULL;
char **hosts = NULL;
char *nodelist;
/* 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 */
env_set_name();
/* if I am a daemon, complete my setup using the
* default procedure
*/
if (ORTE_PROC_IS_DAEMON) {
/* get the list of nodes used for this job */
nodelist = getenv("OMPI_MCA_orte_nodelist");
if (NULL != nodelist) {
/* split the node list into an argv array */
hosts = opal_argv_split(nodelist, ',');
}
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_orted_setup";
goto error;
}
opal_argv_free(hosts);
return ORTE_SUCCESS;
}
if (ORTE_PROC_IS_TOOL) {
/* otherwise, if I am a tool proc, use that procedure */
if (ORTE_SUCCESS != (ret = orte_ess_base_tool_setup())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_tool_setup";
goto error;
}
/* as a tool, I don't need a nidmap - so just return now */
return ORTE_SUCCESS;
}
/* otherwise, I must be an application process - 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;
}
/* if one was provided, build my nidmap */
if (ORTE_SUCCESS != (ret = orte_util_nidmap_init(orte_process_info.sync_buf))) {
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;
/* if I am a daemon, finalize using the default procedure */
if (ORTE_PROC_IS_DAEMON) {
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_finalize())) {
ORTE_ERROR_LOG(ret);
}
} else if (ORTE_PROC_IS_TOOL) {
/* otherwise, if I am a tool proc, use that procedure */
if (ORTE_SUCCESS != (ret = orte_ess_base_tool_finalize())) {
ORTE_ERROR_LOG(ret);
}
/* as a tool, I didn't create a nidmap - so just return now */
return ret;
} else {
/* otherwise, I must be an application process
* 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 uint8_t proc_get_locality(orte_process_name_t *proc)
{
orte_nid_t *nid;
if (NULL == (nid = orte_util_lookup_nid(proc))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return OPAL_PROC_NON_LOCAL;
}
if (nid->daemon == ORTE_PROC_MY_DAEMON->vpid) {
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s on LOCAL NODE",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc)));
return (OPAL_PROC_ON_NODE | OPAL_PROC_ON_CU | OPAL_PROC_ON_CLUSTER);
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s is REMOTE",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc)));
return OPAL_PROC_NON_LOCAL;
}
static orte_vpid_t proc_get_daemon(orte_process_name_t *proc)
{
orte_nid_t *nid;
if (NULL == (nid = orte_util_lookup_nid(proc))) {
return ORTE_VPID_INVALID;
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s is hosted by daemon %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc),
ORTE_VPID_PRINT(nid->daemon)));
return nid->daemon;
}
static char* proc_get_hostname(orte_process_name_t *proc)
{
orte_nid_t *nid;
if (NULL == (nid = orte_util_lookup_nid(proc))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return NULL;
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s is on host %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc),
nid->name));
return nid->name;
}
static uint32_t proc_get_arch(orte_process_name_t *proc)
{
orte_nid_t *nid;
if (NULL == (nid = orte_util_lookup_nid(proc))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return 0;
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s has arch %0x",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc),
nid->arch));
return nid->arch;
}
static int update_arch(orte_process_name_t *proc, uint32_t arch)
{
orte_nid_t *nid;
if (NULL == (nid = orte_util_lookup_nid(proc))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: updating proc %s to arch %0x",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc),
arch));
nid->arch = arch;
return ORTE_SUCCESS;
}
static orte_local_rank_t proc_get_local_rank(orte_process_name_t *proc)
{
orte_pmap_t *pmap;
if (NULL == (pmap = orte_util_lookup_pmap(proc))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_LOCAL_RANK_INVALID;
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s has local rank %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc),
(int)pmap->local_rank));
return pmap->local_rank;
}
static orte_node_rank_t proc_get_node_rank(orte_process_name_t *proc)
{
orte_pmap_t *pmap;
/* is this me? */
if (proc->jobid == ORTE_PROC_MY_NAME->jobid &&
proc->vpid == ORTE_PROC_MY_NAME->vpid) {
/* yes it is - reply with my rank. This is necessary
* because the pidmap will not have arrived when I
* am starting up, and if we use static ports, then
* I need to know my node rank during init
*/
return my_node_rank;
}
if (NULL == (pmap = orte_util_lookup_pmap(proc))) {
return ORTE_NODE_RANK_INVALID;
}
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: proc %s has node rank %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(proc),
(int)pmap->node_rank));
return pmap->node_rank;
}
static int update_pidmap(opal_byte_object_t *bo)
{
int ret;
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_output,
"%s ess:env: updating pidmap",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* build the pmap */
if (ORTE_SUCCESS != (ret = orte_util_decode_pidmap(bo))) {
ORTE_ERROR_LOG(ret);
}
return ret;
}
static int update_nidmap(opal_byte_object_t *bo)
{
int rc;
/* decode the nidmap - the util will know what to do */
if (ORTE_SUCCESS != (rc = orte_util_decode_nodemap(bo))) {
ORTE_ERROR_LOG(rc);
}
return rc;
}
static int env_set_name(void)
{
char *tmp;
int rc;
orte_jobid_t jobid;
orte_vpid_t vpid;
mca_base_param_reg_string_name("orte", "ess_jobid", "Process jobid",
true, false, NULL, &tmp);
if (NULL == tmp) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, tmp))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
free(tmp);
mca_base_param_reg_string_name("orte", "ess_vpid", "Process vpid",
true, false, NULL, &tmp);
if (NULL == tmp) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, tmp))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
free(tmp);
ORTE_PROC_MY_NAME->jobid = jobid;
ORTE_PROC_MY_NAME->vpid = vpid;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:env set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* get my node rank in case we are using static ports - this won't
* be present for daemons, so don't error out if we don't have it
*/
mca_base_param_reg_string_name("orte", "ess_node_rank", "Process node rank",
true, false, NULL, &tmp);
if (NULL != tmp) {
my_node_rank = strtol(tmp, NULL, 10);
}
/* 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;
orte_proc_type_t svtype;
/******** 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 = env_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.
*/
svtype = orte_process_info.proc_type;
orte_process_info.proc_type = ORTE_PROC_TOOL;
if (ORTE_SUCCESS != (ret = orte_routed.finalize()) ) {
exit_status = ret;
goto cleanup;
}
orte_process_info.proc_type = svtype;
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_env_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_env_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