[ovirt-users] Re: Ovirt 4.5 HA over NFS fails when a single host goes down

You must not have an even number of HE-capable nodes. What you're running into is a classic split-brain scenario, with only two nodes allowed to run the HE, and one of the nodes down, the surviving host does not have quorum so does not know it can safely power off the other machine (because obviously this surviving node, from its viewpoint, may have somehow become isolated from the network while the other host is happily alive and running the engine and controlling everything). In clustering, you _never_ want two, or four, or six of something. One, three, five, etc... because it must be impossible to have a "tie" situation when the decisions are being made on which hosts are needing to be fenced.