HA High Availability Failed to move VMs

[LesCrisps]

Hello. We had what we thought was a properly configured HA cluster but it didn't work as expected during a recent host failure. We have a stretched single cluster with two datacentres and a fully supported HP quorum-based SAN setup. We are on on 5.5 and vCenter is on the latest update and the hosts are on the last but one update.

The VMs have a soft-affinity for the hosts nearest to their SAN so would failover to any host in DR.

I don't think any of the above setup is part of the issue, I think HA itself had a problem. All hosts were connected and there were no configuration issues.

A blade was pulled from its chassis but the VMs weren't restarted elsewhere. They showed as 'Disconnected' in the client and were unable to be restarted manually until the blade was connected again.

Has anybody had experience of this?

[bheemeswararao]

similar issue at my end. My HA failed becuase of the network setups that we made incorrect in vMotion network. Mostly it would be the same case with you. Check your network setting on both nodes and it should match it.

I am assuming that vMotion is working fine on your cluster.

[LesCrisps]

Hi. Yes, vMotion works fine within and across datacentres. Could you go into more detail please. If we've never noticed a problem with vMotion how could this be a factor and can you define what you mean by 'both nodes'.

The HA restarts should have stayed within the Datacentre as we have two chassis with four esxi blades in each and plenty of spare capacity. Soft-affinity rules should have placed them near-by.

Either HA wasn't aware of the host failure, or it was and couldn't do anything about it.