File pacemaker-pengine-dont-keep-unique-instances-on... of Package pacemaker.13547

Overview Repositories Revisions Requests Users Attributes Meta

File pacemaker-pengine-dont-keep-unique-instances-on-same-node.patch of Package pacemaker.13547

commit 6c4aa5023b8c985cca9d7137e5f7de41015cf282
Author: Ken Gaillot <kgaillot@redhat.com>
Date:   Tue Oct 24 15:37:00 2017 -0500

Fix: pengine: don't keep unique instances on same node
    
    Fixes CLBZ#5326

Index: pacemaker/pengine/native.c
===================================================================
--- pacemaker.orig/pengine/native.c
+++ pacemaker/pengine/native.c
@@ -148,9 +148,18 @@ native_choose_node(resource_t * rsc, nod
         pe_rsc_trace(rsc, "Chose node %s for %s from %d candidates",
                      chosen ? chosen->details->uname : "<none>", rsc->id, length);
 
-        if (chosen && chosen->weight > 0 && can_run_resources(chosen)) {
+        if (!((rsc->parent
+               && ((rsc->parent->variant == pe_clone) || (rsc->parent->variant == pe_master)))
+              && is_set(rsc->parent->flags, pe_rsc_unique))
+            && chosen && (chosen->weight > 0) && can_run_resources(chosen)) {
             /* If the resource is already running on a node, prefer that node if
              * it is just as good as the chosen node.
+             *
+             * We don't do this for unique clone instances, because
+             * distribute_children() has already assigned instances to their
+             * running nodes when appropriate, and if we get here, we don't want
+             * remaining unallocated instances to prefer a node that's already
+             * running another instance.
              */
             node_t *running = g_list_nth_data(rsc->running_on, 0);

Places

File pacemaker-pengine-dont-keep-unique-instances-on-same-node.patch of Package pacemaker.13547

Places