Merge pull request #4232 from rhc54/topic/local

Implement support for "local" range when publishing data

opal/mca/pmix/pmix2x/pmix/examples/pub.c

@@ -41,6 +41,7 @@ int main(int argc, char **argv)
     uint32_t nprocs;
     pmix_info_t *info;
     pmix_pdata_t *pdata;
+    pmix_data_range_t range = PMIX_RANGE_LOCAL;
 
     /* init us */
     if (PMIX_SUCCESS != (rc = PMIx_Init(&myproc, NULL, 0))) {
@@ -69,14 +70,15 @@ int main(int argc, char **argv)
 
     /* publish something */
     if (0 == myproc.rank) {
-        PMIX_INFO_CREATE(info, 2);
+        PMIX_INFO_CREATE(info, 3);
         (void)strncpy(info[0].key, "FOOBAR", PMIX_MAX_KEYLEN);
         info[0].value.type = PMIX_UINT8;
         info[0].value.data.uint8 = 1;
         (void)strncpy(info[1].key, "PANDA", PMIX_MAX_KEYLEN);
         info[1].value.type = PMIX_SIZE;
         info[1].value.data.size = 123456;
-        if (PMIX_SUCCESS != (rc = PMIx_Publish(info, 2))) {
+        PMIX_INFO_LOAD(&info[2], PMIX_RANGE, &range, PMIX_DATA_RANGE);
+        if (PMIX_SUCCESS != (rc = PMIx_Publish(info, 3))) {
             fprintf(stderr, "Client ns %s rank %d: PMIx_Publish failed: %d\n", myproc.nspace, myproc.rank, rc);
             goto done;
         }

orte/mca/ess/hnp/ess_hnp_module.c

@@ -96,7 +96,6 @@
 #include "orte/runtime/orte_quit.h"
 #include "orte/runtime/orte_cr.h"
 #include "orte/runtime/orte_locks.h"
-#include "orte/runtime/orte_data_server.h"
 
 #include "orte/mca/ess/ess.h"
 #include "orte/mca/ess/base/base.h"
@@ -622,13 +621,6 @@ static int rte_init(void)
     orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_SHOW_HELP,
                             ORTE_RML_PERSISTENT, orte_show_help_recv, NULL);
 
-    /* setup the data server */
-    if (ORTE_SUCCESS != (ret = orte_data_server_init())) {
-        ORTE_ERROR_LOG(ret);
-        error = "orte_data_server_init";
-        goto error;
-    }
-
     if (orte_create_session_dirs) {
         /* set the opal_output hnp file location to be in the
          * proc-specific session directory. */
@@ -817,9 +809,6 @@ static int rte_finalize(void)
     /* shutdown the pmix server */
     pmix_server_finalize();
     (void) mca_base_framework_close(&opal_pmix_base_framework);
-    /* cleanup our data server */
-    orte_data_server_finalize();
-
     (void) mca_base_framework_close(&orte_dfs_base_framework);
     (void) mca_base_framework_close(&orte_filem_base_framework);
     /* output any lingering stdout/err data */
@@ -921,8 +910,8 @@ static void clean_abort(int fd, short flags, void *arg)
             orte_odls.kill_local_procs(NULL);
             /* whack any lingering session directory files from our jobs */
             orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
-            /* cleanup our data server */
-            orte_data_server_finalize();
+            /* cleanup our pmix server */
+            opal_pmix.finalize();
             /* exit with a non-zero status */
             exit(ORTE_ERROR_DEFAULT_EXIT_CODE);
         }
@@ -941,10 +930,6 @@ static void clean_abort(int fd, short flags, void *arg)
      * so need to tell them that!
      */
     orte_execute_quiet = true;
-    if (!orte_never_launched) {
-        /* cleanup our data server */
-        orte_data_server_finalize();
-    }
     /* We are in an event handler; the job completed procedure
        will delete the signal handler that is currently running
        (which is a Bad Thing), so we can't call it directly.

orte/orted/pmix/pmix_server.c

@@ -70,6 +70,7 @@
 #include "orte/util/show_help.h"
 #include "orte/util/threads.h"
 #include "orte/runtime/orte_globals.h"
+#include "orte/runtime/orte_data_server.h"
 
 #include "pmix_server.h"
 #include "pmix_server_internal.h"
@@ -293,6 +294,9 @@ int pmix_server_init(void)
 
 void pmix_server_start(void)
 {
+    /* setup our local data server */
+    orte_data_server_init();
+
     /* setup recv for direct modex requests */
      orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DIRECT_MODEX,
                              ORTE_RML_PERSISTENT, pmix_server_dmdx_recv, NULL);
@@ -331,6 +335,9 @@ void pmix_server_finalize(void)
     orte_rml.recv_cancel(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DATA_CLIENT);
     orte_rml.recv_cancel(ORTE_NAME_WILDCARD, ORTE_RML_TAG_NOTIFICATION);
 
+    /* finalize our local data server */
+    orte_data_server_finalize();
+
     /* shutdown the local server */
     opal_pmix.server_finalize();
 

orte/orted/pmix/pmix_server_pub.c

@@ -187,8 +187,20 @@ static void execute(int sd, short args, void *cbdata)
 
     /* if the range is SESSION, then set the target to the global server */
     if (OPAL_PMIX_RANGE_SESSION == req->range) {
+        opal_output_verbose(1, orte_pmix_server_globals.output,
+                            "%s orted:pmix:server range SESSION",
+                            ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
         target = &orte_pmix_server_globals.server;
+    } else if (OPAL_PMIX_RANGE_LOCAL == req->range) {
+        /* if the range is local, send it to myself */
+        opal_output_verbose(1, orte_pmix_server_globals.output,
+                            "%s orted:pmix:server range LOCAL",
+                            ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
+        target = ORTE_PROC_MY_NAME;
     } else {
+        opal_output_verbose(1, orte_pmix_server_globals.output,
+                            "%s orted:pmix:server range GLOBAL",
+                            ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
         target = ORTE_PROC_MY_HNP;
     }