From e1aa7939efe635757dbaef18115e8ab6808bd1f5 Mon Sep 17 00:00:00 2001
From: Ralph Castain <rhc@open-mpi.org>
Date: Wed, 7 Dec 2016 13:53:00 -0800
Subject: [PATCH] Correctly cleanup the local children and node map info on
 remote orteds upon job completion. Ensure that register_nspace only includes
 procs from that job in the proc map

Thanks to Ashley Pittman for the report

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
---
 orte/mca/state/orted/state_orted.c         | 27 +++++++++++++++++++++-
 orte/orted/pmix/pmix_server_register_fns.c |  4 ++++
 2 files changed, 30 insertions(+), 1 deletion(-)

diff --git a/orte/mca/state/orted/state_orted.c b/orte/mca/state/orted/state_orted.c
index d77d6ec178..214e223cf2 100644
--- a/orte/mca/state/orted/state_orted.c
+++ b/orte/mca/state/orted/state_orted.c
@@ -252,7 +252,7 @@ static void track_procs(int fd, short argc, void *cbdata)
     orte_job_t *jdata;
     orte_proc_t *pdata, *pptr;
     opal_buffer_t *alert;
-    int rc, i;
+    int rc, i, j;
     orte_plm_cmd_flag_t cmd;
     char *rtmod;
 
@@ -416,6 +416,31 @@ static void track_procs(int fd, short argc, void *cbdata)
             }
             /* mark that we sent it so we ensure we don't do it again */
             orte_set_attribute(&jdata->attributes, ORTE_JOB_TERM_NOTIFIED, ORTE_ATTR_LOCAL, NULL, OPAL_BOOL);
+            /* cleanup the procs as these are gone */
+            for (i=0; i < orte_local_children->size; i++) {
+                if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(orte_local_children, i))) {
+                    continue;
+                }
+                /* if this child is part of the job... */
+                if (pptr->name.jobid == jdata->jobid) {
+                    /* clear the entry in the local children */
+                    opal_pointer_array_set_item(orte_local_children, i, NULL);
+                    /* find it in the node->procs array */
+                    for (j=0; j < pptr->node->procs->size; j++) {
+                        if (NULL == (pdata = (orte_proc_t*)opal_pointer_array_get_item(pptr->node->procs, j))) {
+                            continue;
+                        }
+                        if (pdata == pptr) {
+                            /* remove it */
+                            opal_pointer_array_set_item(pptr->node->procs, j, NULL);
+                            OBJ_RELEASE(pdata);  // maintain accounting
+                            break;
+                        }
+                    }
+                    OBJ_RELEASE(pptr);  // maintain accounting
+                }
+            }
+
         }
     }
 
diff --git a/orte/orted/pmix/pmix_server_register_fns.c b/orte/orted/pmix/pmix_server_register_fns.c
index f53f8feb5c..3d0cf381e0 100644
--- a/orte/orted/pmix/pmix_server_register_fns.c
+++ b/orte/orted/pmix/pmix_server_register_fns.c
@@ -332,6 +332,10 @@ int orte_pmix_server_register_nspace(orte_job_t *jdata)
             if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(node->procs, i))) {
                 continue;
             }
+            /* only consider procs from this job */
+            if (pptr->name.jobid != jdata->jobid) {
+                continue;
+            }
             /* setup the proc map object */
             kv = OBJ_NEW(opal_value_t);
             kv->key = strdup(OPAL_PMIX_PROC_DATA);