root/orte/mca/ess/slurm/ess_slurm_module.c

DEFINITIONS

1. rte_init
2. rte_finalize
3. slurm_set_name

   1 /*
   2  * Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
   3  *                         University Research and Technology
   4  *                         Corporation.  All rights reserved.
   5  * Copyright (c) 2004-2011 The University of Tennessee and The University
   6  *                         of Tennessee Research Foundation.  All rights
   7  *                         reserved.
   8  * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
   9  *                         University of Stuttgart.  All rights reserved.
  10  * Copyright (c) 2004-2005 The Regents of the University of California.
  11  *                         All rights reserved.
  12  * Copyright (c) 2008-2011 Cisco Systems, Inc.  All rights reserved.
  13  * Copyright (c) 2013-2018 Intel, Inc.  All rights reserved.
  14  * $COPYRIGHT$
  15  *
  16  * Additional copyrights may follow
  17  *
  18  * $HEADER$
  19  *
  20  */
  21 
  22 #include "orte_config.h"
  23 #include "orte/constants.h"
  24 
  25 #ifdef HAVE_UNISTD_H
  26 #include <unistd.h>
  27 #endif  /* HAVE_UNISTD_H */
  28 #include <string.h>
  29 #include <ctype.h>
  30 
  31 
  32 #include "opal/util/opal_environ.h"
  33 #include "opal/util/output.h"
  34 #include "opal/util/argv.h"
  35 #include "opal/class/opal_pointer_array.h"
  36 #include "opal/dss/dss.h"
  37 
  38 #include "orte/util/proc_info.h"
  39 #include "orte/util/show_help.h"
  40 #include "orte/mca/errmgr/errmgr.h"
  41 #include "orte/mca/rml/rml.h"
  42 #include "orte/util/name_fns.h"
  43 #include "orte/runtime/orte_globals.h"
  44 
  45 #include "orte/mca/ess/ess.h"
  46 #include "orte/mca/ess/base/base.h"
  47 #include "orte/mca/ess/slurm/ess_slurm.h"
  48 
  49 static int slurm_set_name(void);
  50 
  51 static int rte_init(void);
  52 static int rte_finalize(void);
  53 
  54 orte_ess_base_module_t orte_ess_slurm_module = {
  55     rte_init,
  56     rte_finalize,
  57     NULL,
  58     NULL /* ft_event */
  59 };
  60 
  61 static int rte_init(void)
  62 {
  63     int ret;
  64     char *error = NULL;
  65 
  66     /* run the prolog */
  67     if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
  68         error = "orte_ess_base_std_prolog";
  69         goto error;
  70     }
  71 
  72     /* Start by getting a unique name */
  73     slurm_set_name();
  74 
  75     /* if I am a daemon, complete my setup using the
  76      * default procedure
  77      */
  78     if (ORTE_PROC_IS_DAEMON) {
  79         if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup())) {
  80             ORTE_ERROR_LOG(ret);
  81             error = "orte_ess_base_orted_setup";
  82             goto error;
  83         }
  84         return ORTE_SUCCESS;
  85     }
  86 
  87     if (ORTE_PROC_IS_TOOL) {
  88         /* otherwise, if I am a tool proc, use that procedure */
  89         if (ORTE_SUCCESS != (ret = orte_ess_base_tool_setup(NULL))) {
  90             ORTE_ERROR_LOG(ret);
  91             error = "orte_ess_base_tool_setup";
  92             goto error;
  93         }
  94         return ORTE_SUCCESS;
  95 
  96     }
  97 
  98     /* no other options are supported! */
  99     error = "ess_error";
 100     ret = ORTE_ERROR;
 101 
 102 error:
 103     if (ORTE_ERR_SILENT != ret && !orte_report_silent_errors) {
 104         orte_show_help("help-orte-runtime.txt",
 105                        "orte_init:startup:internal-failure",
 106                        true, error, ORTE_ERROR_NAME(ret), ret);
 107     }
 108 
 109     return ret;
 110 }
 111 
 112 static int rte_finalize(void)
 113 {
 114     int ret;
 115 
 116     /* if I am a daemon, finalize using the default procedure */
 117     if (ORTE_PROC_IS_DAEMON) {
 118         if (ORTE_SUCCESS != (ret = orte_ess_base_orted_finalize())) {
 119             ORTE_ERROR_LOG(ret);
 120             return ret;
 121         }
 122     } else if (ORTE_PROC_IS_TOOL) {
 123         /* otherwise, if I am a tool proc, use that procedure */
 124         if (ORTE_SUCCESS != (ret = orte_ess_base_tool_finalize())) {
 125             ORTE_ERROR_LOG(ret);
 126         }
 127         return ret;
 128     }
 129 
 130     return ORTE_SUCCESS;
 131 }
 132 
 133 static int slurm_set_name(void)
 134 {
 135     int slurm_nodeid;
 136     int rc;
 137     orte_jobid_t jobid;
 138     orte_vpid_t vpid;
 139     char *tmp;
 140 
 141     OPAL_OUTPUT_VERBOSE((1, orte_ess_base_framework.framework_output,
 142                          "ess:slurm setting name"));
 143 
 144     if (NULL == orte_ess_base_jobid) {
 145         ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
 146         return ORTE_ERR_NOT_FOUND;
 147     }
 148     if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, orte_ess_base_jobid))) {
 149         ORTE_ERROR_LOG(rc);
 150         return(rc);
 151     }
 152 
 153     if (NULL == orte_ess_base_vpid) {
 154         ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
 155         return ORTE_ERR_NOT_FOUND;
 156     }
 157     if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, orte_ess_base_vpid))) {
 158         ORTE_ERROR_LOG(rc);
 159         return(rc);
 160     }
 161 
 162     ORTE_PROC_MY_NAME->jobid = jobid;
 163 
 164     /* fix up the vpid and make it the "real" vpid */
 165     slurm_nodeid = atoi(getenv("SLURM_NODEID"));
 166     ORTE_PROC_MY_NAME->vpid = vpid + slurm_nodeid;
 167 
 168     OPAL_OUTPUT_VERBOSE((1, orte_ess_base_framework.framework_output,
 169                          "ess:slurm set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
 170 
 171     /* fix up the system info nodename to match exactly what slurm returned */
 172     if (NULL != orte_process_info.nodename) {
 173         free(orte_process_info.nodename);
 174     }
 175     if (NULL == (tmp = getenv("SLURMD_NODENAME"))) {
 176         ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
 177         return ORTE_ERR_NOT_FOUND;
 178     }
 179     orte_process_info.nodename = strdup(tmp);
 180 
 181 
 182     OPAL_OUTPUT_VERBOSE((1, orte_ess_base_framework.framework_output,
 183                          "ess:slurm set nodename to %s",
 184                          (NULL == orte_process_info.nodename) ? "NULL" : orte_process_info.nodename));
 185 
 186     /* get the non-name common environmental variables */
 187     if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
 188         ORTE_ERROR_LOG(rc);
 189         return rc;
 190     }
 191 
 192     return ORTE_SUCCESS;
 193 }