Description of problem: InstanceHA feature rebuilds the instances only in case of compute node crash, can we extend the feature for node power cycle also? Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Hello, Thanks for the RFE. I think our PIDONE team, responsible for InstanceHA, is best placed to give you a definitive answer on this, so I've re-assigned this BZ to them. In the meantime, I'd like to pre-emptively ask you for more information - specifically, what you mean by "node power cycle"? Could you describe in detail how you see the new instance HA behaviour working, for example: 1. I manually power off the compute node. 2. I expect all of its instances to be evacuated to other hosts. Also, if you could explain *why* you need this new behaviour, we might be able to guide you to a different solution, in case the specific RFE you're asking for is refused for whatever reason. Thanks!
Hello, Yes scheduled/unscheduled power shutdown is a potential scenario when we will need the automatic evacuation of instances to other hosts.
Instance HA is exclusively targeted at resurrecting dead instances in the event of a compute node crash/failure. Clean shutdowns, whether planned or not, are very much out of scope for this feature and are better handled by live migration prior to shutdown.
Can you please elaborate on what are the events classified as a "compute node crash/failure" for InstanceHA to be triggered?
Anything that prevents the pacemaker cluster from communicating with its peer process on the compute node. Eg. network or power loss Recovering from failures of individual instances is explicitly out of scope.