Add missing functionaltiy to the ns replica to support remote get_job_peers requests. Add trace commands to help try and track down remaining problem with comm_spawn.

This commit was SVN r11939.

orte/mca/ns/proxy/src/ns_proxy.c

@@ -25,8 +25,11 @@
 
 #include "orte/orte_constants.h"
 #include "orte/orte_types.h"
+
 #include "opal/mca/mca.h"
 #include "opal/util/output.h"
+#include "opal/util/trace.h"
+
 #include "orte/dss/dss.h"
 #include "orte/mca/errmgr/errmgr.h"
 #include "orte/mca/rml/rml.h"
@@ -50,6 +53,8 @@ int orte_ns_proxy_create_cellid(orte_cellid_t *cellid, char *site, char *resourc
     int rc;
     orte_ns_proxy_cell_info_t *new_cell;
 
+    OPAL_TRACE(1);
+    
     /* set the default value of error */
     *cellid = ORTE_CELLID_MAX;
 
@@ -160,6 +165,8 @@ int orte_ns_proxy_get_cell_info(orte_cellid_t cellid,
     orte_ns_proxy_cell_info_t **cell, *new_cell;
     int rc, ret=ORTE_SUCCESS;
 
+    OPAL_TRACE(1);
+    
     /* see if we already have the info locally */
     OPAL_THREAD_LOCK(&orte_ns_proxy.mutex);
 
@@ -299,6 +306,8 @@ int orte_ns_proxy_create_jobid(orte_jobid_t *job)
     orte_std_cntr_t count;
     int rc;
 
+    OPAL_TRACE(1);
+    
     /* set default value */
     *job = ORTE_JOBID_MAX;
 
@@ -365,6 +374,8 @@ int orte_ns_proxy_reserve_range(orte_jobid_t job, orte_vpid_t range, orte_vpid_t
     orte_std_cntr_t count;
     int rc;
 
+    OPAL_TRACE(1);
+    
     /* set default return value */
     *starting_vpid = ORTE_VPID_MAX;
 
@@ -439,9 +450,11 @@ int orte_ns_proxy_get_job_peers(orte_process_name_t **procs,
     orte_buffer_t* cmd;
     orte_buffer_t* answer;
     orte_ns_cmd_flag_t command;
-    orte_std_cntr_t count;
+    orte_std_cntr_t count, nprocs;
     int rc;
 
+    OPAL_TRACE_ARG1(1, job);
+    
     /* set default value */
     *procs = NULL;
     *num_procs = 0;
@@ -452,12 +465,20 @@ int orte_ns_proxy_get_job_peers(orte_process_name_t **procs,
     }
 
     command = ORTE_NS_GET_JOB_PEERS_CMD;
+    /* pack the command */
     if (ORTE_SUCCESS != (rc = orte_dss.pack(cmd, (void*)&command, 1, ORTE_NS_CMD))) { /* got a problem */
         ORTE_ERROR_LOG(rc);
         OBJ_RELEASE(cmd);
         return rc;
     }
 
+    /* pack the jobid */
+    if (ORTE_SUCCESS != (rc = orte_dss.pack(cmd, &job, 1, ORTE_JOBID))) {
+        ORTE_ERROR_LOG(rc);
+        OBJ_RELEASE(cmd);
+        return rc;
+    }
+
     if (0 > orte_rml.send_buffer(orte_ns_proxy.my_replica, cmd, ORTE_RML_TAG_NS, 0)) {
         ORTE_ERROR_LOG(ORTE_ERR_COMM_FAILURE);
         OBJ_RELEASE(cmd);
@@ -491,27 +512,28 @@ int orte_ns_proxy_get_job_peers(orte_process_name_t **procs,
     }
 
     count = 1;
-    if (ORTE_SUCCESS != (rc = orte_dss.unpack(answer, &num_procs, &count, ORTE_STD_CNTR))) {
+    if (ORTE_SUCCESS != (rc = orte_dss.unpack(answer, &nprocs, &count, ORTE_STD_CNTR))) {
         ORTE_ERROR_LOG(rc);
         OBJ_RELEASE(answer);
         return rc;
     }
 
     /* allocate space for array of proc names */
-    if (0 < *num_procs) {
+    if (0 < nprocs) {
-        *procs = (orte_process_name_t*)malloc((*num_procs) * sizeof(orte_process_name_t));
+        *procs = (orte_process_name_t*)malloc((nprocs) * sizeof(orte_process_name_t));
         if (NULL == *procs) {
             ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
             OBJ_RELEASE(answer);
             return ORTE_ERR_OUT_OF_RESOURCE;
         }
 
-        if (ORTE_SUCCESS != (rc = orte_dss.unpack(answer, procs, num_procs, ORTE_NAME))) {
+        if (ORTE_SUCCESS != (rc = orte_dss.unpack(answer, procs, &nprocs, ORTE_NAME))) {
             ORTE_ERROR_LOG(rc);
             OBJ_RELEASE(answer);
             return rc;
         }
     }
+    *num_procs = nprocs;
 
     OBJ_RELEASE(answer);
     return ORTE_SUCCESS;

orte/mca/ns/replica/src/ns_replica.c

@@ -23,11 +23,13 @@
 #include <stdio.h>
 #include <string.h>
 
-#include "orte/dss/dss.h"
 #include "opal/threads/mutex.h"
-
 #include "opal/util/output.h"
+#include "opal/util/trace.h"
+
+#include "orte/dss/dss.h"
 #include "orte/mca/errmgr/errmgr.h"
+
 #include "orte/mca/ns/base/base.h"
 #include "ns_replica.h"
 
@@ -46,6 +48,8 @@ int orte_ns_replica_create_cellid(orte_cellid_t *cellid, char *site, char *resou
     int rc;
     orte_std_cntr_t index;
 
+    OPAL_TRACE(1);
+    
     OPAL_THREAD_LOCK(&orte_ns_replica.mutex);
 
     *cellid = ORTE_CELLID_MAX;
@@ -90,6 +94,8 @@ int orte_ns_replica_get_cell_info(orte_cellid_t cellid,
     orte_cellid_t j;
     orte_ns_replica_cell_tracker_t **cell;
 
+    OPAL_TRACE(1);
+    
     OPAL_THREAD_LOCK(&orte_ns_replica.mutex);
 
     cell = (orte_ns_replica_cell_tracker_t**)(orte_ns_replica.cells)->addr;
@@ -120,6 +126,8 @@ int orte_ns_replica_create_jobid(orte_jobid_t *jobid)
     int rc;
     orte_std_cntr_t index;
 
+    OPAL_TRACE(1);
+    
     OPAL_THREAD_LOCK(&orte_ns_replica.mutex);
 
     *jobid = ORTE_JOBID_MAX;
@@ -164,6 +172,8 @@ int orte_ns_replica_reserve_range(orte_jobid_t job, orte_vpid_t range,
     orte_std_cntr_t j;
     orte_jobid_t k;
 
+    OPAL_TRACE(1);
+    
     OPAL_THREAD_LOCK(&orte_ns_replica.mutex);
 
     /* find the jobid */
@@ -204,6 +214,8 @@ int orte_ns_replica_get_job_peers(orte_process_name_t **procs,
     orte_std_cntr_t j;
     orte_jobid_t k;
 
+    OPAL_TRACE_ARG1(1, job);
+    
     OPAL_THREAD_LOCK(&orte_ns_replica.mutex);
 
     /* find the jobid */

orte/mca/ns/replica/src/ns_replica_component.c

@@ -412,7 +412,8 @@ void orte_ns_replica_recv(int status, orte_process_name_t* sender,
     char *tagname, *site, *resource;
     orte_rml_tag_t oob_tag;
     orte_data_type_t type;
-    orte_std_cntr_t count;
+    orte_std_cntr_t count, nprocs;
+    orte_process_name_t *procs;
     int rc=ORTE_SUCCESS, ret;
 
     count = 1;
@@ -611,6 +612,33 @@ void orte_ns_replica_recv(int status, orte_process_name_t* sender,
             /* ignore this command */
             break;
 
+        case ORTE_NS_GET_JOB_PEERS_CMD:
+            /* unpack the jobid */
+            count = 1;
+            if (ORTE_SUCCESS != (rc = orte_dss.unpack(buffer, &job, &count, ORTE_JOBID))) {
+                ORTE_ERROR_LOG(rc);
+                goto RETURN_ERROR;
+            }
+            /* process the request */
+            if (ORTE_SUCCESS != (rc = orte_ns_replica_get_job_peers(&procs, &nprocs, job))) {
+                ORTE_ERROR_LOG(rc);
+                goto RETURN_ERROR;
+            }
+                
+            /* pack the answer */
+            if (ORTE_SUCCESS != (rc = orte_dss.pack(&answer, &nprocs, 1, ORTE_STD_CNTR))) {
+                ORTE_ERROR_LOG(rc);
+                goto RETURN_ERROR;
+            }
+                
+            if (nprocs > 0) {
+                if (ORTE_SUCCESS != (rc = orte_dss.pack(&answer, &procs, nprocs, ORTE_NAME))) {
+                    ORTE_ERROR_LOG(rc);
+                    goto RETURN_ERROR;
+                }
+            }
+            break;
+            
         case ORTE_NS_DUMP_CELLS_CMD:
             if (ORTE_SUCCESS != (rc = orte_ns_replica_dump_cells_fn(&answer))) {
                 ORTE_ERROR_LOG(rc);

orte/mca/rmgr/base/rmgr_base_stage_gate.c

@@ -85,6 +85,8 @@ int orte_rmgr_base_proc_stage_gate_mgr(orte_gpr_notify_message_t *msg)
          return rc;
      }
 
+    OPAL_TRACE_ARG1(1, job);
+    
     /* need the list of peers for this job so we can send them the xcast.
      * obtain this list from the name service's get_job_peers function
      */