Issue

• We issued a pcs resource restart on our openstack-nova-api resource and all the compute nodes are fenced by Pacemaker and all the VM instances stopped.
• The nova-compute resource is regularly timing out on stop during a resource restart or cluster node shutdown, causing controller node fencing
• nova-compute fails with OCF_TIMEOUT
• Our nova-compute resource is failing to stop due to a timeout error

  Sep 03 15:25:03 [3480] node1    pengine:  warning: unpack_rsc_op_failure: Processing failed op stop for nova-compute:0 on compute-l01: OCF_TIMEOUT (198)
  Sep 03 15:25:03 [3480] node1    pengine:  warning: pe_fence_node: Node compute1 will be fenced because of resource failure(s)

  Sep 01 16:46:44 [3485] node1       crmd:     info: process_graph_event:   Detected action (296.568) nova-compute_stop_0.73608=OCF_TIMEOUT: failed
  Sep 01 16:46:44 [3485] node1       crmd:  warning: status_from_rc:        Action 568 (nova-compute_stop_0) on compute-l02 failed (target: 0 vs. rc: 198): Error
  Sep 01 16:46:44 [3485] node1       crmd:     info: abort_transition_graph:        Transition aborted by nova-compute_stop_0 'modify' on node2: Event failed (magic=2:198;568:296:0:3dd776e6-e59b-4d8b-b1b8-fbce832e096f, cib=0.316.2, source=match_graph_event:381, 0)
  Sep 01 16:46:44 [3485] node1       crmd:     info: match_graph_event:     Action nova-compute_stop_0 (568) confirmed on compute-l02 (rc=198)
  Sep 01 16:46:44 [3485] node1       crmd:     info: update_failcount:      Updating failcount for nova-compute on compute-l02 after failed stop: rc=198 (update=INFINITY, time=1472716004)
  Sep 01 16:46:44 [3485] node1       crmd:     info: process_graph_event:   Detected action (296.568) nova-compute_stop_0.73608=OCF_TIMEOUT: failed