Add some more detailed error output to the db_hash component and nidmap code. Ensure the local nodename is included in the HNP's aliases

This commit was SVN r27622.

orte/include/orte/constants.h

@@ -123,7 +123,10 @@ enum {
     ORTE_ERR_COMM_DISABLED                  = (ORTE_ERR_BASE - 40),
     ORTE_ERR_FAILED_TO_MAP                  = (ORTE_ERR_BASE - 41),
     ORTE_ERR_TAKE_NEXT_OPTION               = (ORTE_ERR_BASE - 42),
-    ORTE_ERR_SENSOR_LIMIT_EXCEEDED          = (ORTE_ERR_BASE - 43)
+    ORTE_ERR_SENSOR_LIMIT_EXCEEDED          = (ORTE_ERR_BASE - 43),
+    ORTE_ERR_JOB_ENTRY_NOT_FOUND            = (ORTE_ERR_BASE - 44),
+    ORTE_ERR_PROC_ENTRY_NOT_FOUND           = (ORTE_ERR_BASE - 45),
+    ORTE_ERR_DATA_VALUE_NOT_FOUND           = (ORTE_ERR_BASE - 46)
 };
 
 #define ORTE_ERR_MAX                      (ORTE_ERR_BASE - 100)

orte/mca/db/hash/db_hash.c

@@ -383,20 +383,20 @@ static int fetch(const orte_process_name_t *proc,
         /* eventually, we will fetch this data - but for now, this
          * is simply an error
          */
-        return ORTE_ERR_NOT_FOUND;
+        return ORTE_ERR_JOB_ENTRY_NOT_FOUND;
     }
 
     /* lookup the proc data object for this proc */
     if (NULL == (proc_data = lookup_orte_proc(jtable->data, proc->vpid))) {
         /* unrecoverable error */
-        return ORTE_ERR_OUT_OF_RESOURCE;
+        return ORTE_ERR_PROC_ENTRY_NOT_FOUND;
     }
 
     /* find the value */
     if (NULL == (kv = lookup_keyval(proc_data, key))) {
         /* again, we eventually will attempt to fetch the data - for
          * now, just report it as an error */
-        return ORTE_ERR_NOT_FOUND;
+        return ORTE_ERR_DATA_VALUE_NOT_FOUND;
     }
 
     /* do the copy and check the type */

orte/mca/ess/hnp/ess_hnp_module.c

@@ -553,6 +553,8 @@ static int rte_init(void)
     /* if we are to retain aliases, get ours */
     if (orte_retain_aliases) {
         opal_ifgetaliases(&node->alias);
+        /* add our own local name to it */
+        opal_argv_append_nosize(&node->alias, orte_process_info.nodename);
     }
 
     /* record that the daemon job is running */

orte/util/error_strings.c

@@ -184,6 +184,15 @@ int orte_err2str(int errnum, const char **errmsg)
     case ORTE_ERR_SENSOR_LIMIT_EXCEEDED:
         retval = "Sensor limit exceeded";
         break;
+    case ORTE_ERR_JOB_ENTRY_NOT_FOUND:
+        retval = "Job entry not found";
+        break;
+    case ORTE_ERR_PROC_ENTRY_NOT_FOUND:
+        retval = "Proc entry not found";
+        break;
+    case ORTE_ERR_DATA_VALUE_NOT_FOUND:
+        retval = "Data not found";
+        break;
     default:
         if (orte_report_silent_errors) {
             retval = "Unknown error";

orte/util/nidmap.c

@@ -383,6 +383,10 @@ int orte_util_decode_nodemap(opal_byte_object_t *bo)
             return rc;
         }
         /* now store a direct reference so we can quickly lookup the daemon from a hostname */
+        opal_output_verbose(2, orte_nidmap_output,
+                            "%s storing nodename %s for daemon %s",
+                            ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                            nodename, ORTE_VPID_PRINT(daemon.vpid));
         if (ORTE_SUCCESS != (rc = orte_db.store(ORTE_NAME_WILDCARD, nodename, &daemon.vpid, ORTE_VPID))) {
             ORTE_ERROR_LOG(rc);
             return rc;
@@ -421,6 +425,10 @@ int orte_util_decode_nodemap(opal_byte_object_t *bo)
                     return rc;
                 }
                 /* store a cross-reference to the daemon for this nodename */
+                opal_output_verbose(2, orte_nidmap_output,
+                                    "%s storing alias %s for daemon %s",
+                                    ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                                    alias, ORTE_VPID_PRINT(daemon.vpid));
                 if (ORTE_SUCCESS != (rc = orte_db.store(ORTE_NAME_WILDCARD, alias, &daemon.vpid, ORTE_VPID))) {
                     ORTE_ERROR_LOG(rc);
                     return rc;