File pacemaker-pengine-dont-clear-same-fail-count-twice.patch of Package pacemaker.14737

commit 357cd703e99fbcf1a371f34966accaf5322b1c50
Author: Ken Gaillot <kgaillot@redhat.com>
Date:   Wed Feb 22 14:14:48 2017 -0600

    Fix: pengine,libpe_status: don't clear same fail-count twice
    
    Previously, pengine and libpe_status were inconsistent when generating
    a key to use for a fail-count op. This could lead to two identical ops
    being scheduled, one of which would timeout (during which time the resource
    would not be recovered). Now, they consistently use generate_op_key().

diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
index 67372739e..a35764392 100644
--- a/lib/pengine/unpack.c
+++ b/lib/pengine/unpack.c
@@ -2934,10 +2934,10 @@ static bool check_operation_expiry(resource_t *rsc, node_t *node, int rc, xmlNod
     }
 
     if (clear_failcount) {
-        action_t *clear_op = NULL;
+        char *key = generate_op_key(rsc->id, CRM_OP_CLEAR_FAILCOUNT, 0);
+        action_t *clear_op = custom_action(rsc, key, CRM_OP_CLEAR_FAILCOUNT,
+                                           node, FALSE, TRUE, data_set);
 
-        clear_op = custom_action(rsc, crm_concat(rsc->id, CRM_OP_CLEAR_FAILCOUNT, '_'),
-                                 CRM_OP_CLEAR_FAILCOUNT, node, FALSE, TRUE, data_set);
         add_hash_param(clear_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE);
     }
 
diff --git a/pengine/allocate.c b/pengine/allocate.c
index 74b57fbb1..4e8d68df8 100644
--- a/pengine/allocate.c
+++ b/pengine/allocate.c
@@ -596,7 +596,7 @@ static gboolean
 failcount_clear_action_exists(node_t * node, resource_t * rsc)
 {
     gboolean rc = FALSE;
-    char *key = crm_concat(rsc->id, CRM_OP_CLEAR_FAILCOUNT, '_');
+    char *key = generate_op_key(rsc->id, CRM_OP_CLEAR_FAILCOUNT, 0);
     GListPtr list = find_actions_exact(rsc->actions, key, node);
 
     if (list) {
@@ -1195,10 +1195,9 @@ cleanup_orphans(resource_t * rsc, pe_working_set_t * data_set)
         node_t *node = (node_t *) gIter->data;
 
         if (node->details->online && get_failcount(node, rsc, NULL, data_set)) {
-            action_t *clear_op = NULL;
-
-            clear_op = custom_action(rsc, crm_concat(rsc->id, CRM_OP_CLEAR_FAILCOUNT, '_'),
-                                     CRM_OP_CLEAR_FAILCOUNT, node, FALSE, TRUE, data_set);
+            char *key = generate_op_key(rsc->id, CRM_OP_CLEAR_FAILCOUNT, 0);
+            action_t *clear_op = custom_action(rsc, key, CRM_OP_CLEAR_FAILCOUNT,
+                                               node, FALSE, TRUE, data_set);
 
             add_hash_param(clear_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE);
             pe_rsc_info(rsc, "Clearing failcount (%d) for orphaned resource %s on %s (%s)",