Description of problem: If two of three controller nodes in an HA setup are taken down, the pacemaker status does not properly display that the two nodes are down. Version-Release number of selected component (if applicable): beta 2 code drop How reproducible: Unsure, only tried it once Steps to Reproduce: 1. Deploy an overcloud with 3 controllers 2. Stop two of the controllers 3. run pcs status on the remaining controller Actual results: Pacemaker shows for some services that only one node is down, with the other two up. It properly stops the VIPs however, since quorum has been lost in this case and nothing will work. Expected results: Services are not expected to work in this situation, but it would be expected that pcs status properly shows the two down nodes. Additional info:
Once you take 2 of 3 controllers out, the cluster is invalid and not recoverable.