Comment 0 for bug 1450399

Revision history for this message
Dennis Dmitriev (ddmitriev) wrote :

Scenario:
    1) Create HA cluster (Ubuntu/Centos) : 3 controllers, 1 compute;
    2) Deploy cluster;
    3) Check that OSTF passed successfully, `pcs status` example: http://paste.openstack.org/show/213411/;
    4) Delete any two controllers from the cluster via Nailgun API/UI/CLI
    5) Run OSTF , check `pcs status` on the remaining controller

Expected resuld: OSTF passed on step 5; `pcs status` show that DC has a quorum and there is running VIPs and other resources.

Actual result: `pcs status` on the remaining controller show that the controller has no quorum (example: http://paste.openstack.org/show/213412/):
"Current DC: node-1.test.domain.local (1) - partition WITHOUT quorum"

, and it is waiting for other controllers that have been deleted:
Online: [ node-1.test.domain.local ]
OFFLINE: [ node-2.test.domain.local node-3.test.domain.local ]

As the result: cluster is in the non-operational state because pacemaker stopped all resources on the remaining controller.

Pacemaker should be re-configured after controller deletion.