File pacemaker-libpe_status-recover-after-failed-dem... of Package pacemaker

Overview Repositories Revisions Requests Users Attributes Meta

File pacemaker-libpe_status-recover-after-failed-demote-when-appropriate.patch of Package pacemaker

commit a962eb7f2aca3f764a1ddecf3c415c2c75a86b01
Author: Ken Gaillot <kgaillot@redhat.com>
Date:   Fri Dec 15 14:41:02 2017 -0600

Low: libpe_status: recover after failed demote when appropriate
    
    Previously, the PE would always do a full stop after any failed demote.
    However, if the resource needed recovery, the necessary start wouln't be
    scheduled until the next transition.
    
    Now, we skip the forced stop. It wasn't needed because native_create_actions()
    does the right thing already. If the resource's next role is stopped, it will
    schedule the right actions to get to stopped. If the resource's next role is
    something else, it will force a stop and start for master/slave resources when
    both a current node and a chosen node have been set (indicating recovery).

diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
index 71307790e..ac1784d88 100644
--- a/lib/pengine/unpack.c
+++ b/lib/pengine/unpack.c
@@ -2662,11 +2662,6 @@ unpack_rsc_op_failure(resource_t * rsc, node_t * node, int rc, xmlNode * xml_op,
         rsc->role = RSC_ROLE_MASTER;
 
     } else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
-        /*
-         * staying in role=master ends up putting the PE/TE into a loop
-         * setting role=slave is not dangerous because no master will be
-         * promoted until the failed resource has been fully stopped
-         */
         if (action->on_fail == action_fail_block) {
             rsc->role = RSC_ROLE_MASTER;
             rsc->next_role = RSC_ROLE_STOPPED;
@@ -2675,9 +2670,13 @@ unpack_rsc_op_failure(resource_t * rsc, node_t * node, int rc, xmlNode * xml_op,
             rsc->role = RSC_ROLE_STOPPED;
 
         } else {
-            crm_warn("Forcing %s to stop after a failed demote action", rsc->id);
+            /*
+             * Staying in master role would put the PE/TE into a loop. Setting
+             * slave role is not dangerous because the resource will be stopped
+             * as part of recovery, and any master promotion will be ordered
+             * after that stop.
+             */
             rsc->role = RSC_ROLE_SLAVE;
-            rsc->next_role = RSC_ROLE_STOPPED;
         }
 
     } else if (compare_version("2.0", op_version) > 0 && safe_str_eq(task, CRMD_ACTION_START)) {

Places

File pacemaker-libpe_status-recover-after-failed-demote-when-appropriate.patch of Package pacemaker

Places