OK, so we have three ESX 3.0.1 host setup and running VMs, a Virtual Centre 2.0.1 server and DRS all configured. Currently all is working fine but I am now scratching my head over HA and gateways for pinging to establish isolation.
I have set up a virtual switch with Service Console and VMkernal, Vmotion enabled. This has 2 NICS assigned, nic1 active for Service Console, nic2 standby and the oposite for the VMkernal port group.
Virtual Machine Port Groups are configured with resilience, being plugged into 2 seperate switches also.
My concern is around the ip address used by HA to establish whether a host is isolated or not. If i specify this as being the gateway (default) or the ip address of another device on our network, everything will be fine until that device is powered off or fails. I can quite happily take down my default gateway without affecting internal clients connecting to virtual machines. However if I invoke HA, and then take down this gateway, all my VMs will power down even though clients could still connect to them.
I have thought this through, trying to remove any single points of failure, but keep coming back to this ip address that HA pings. I guess I need to have a clustered switch or something on my network that provides hardware redundancy behind this ip address? - not a VMware problem exactly but one that will affect HA.
Has anyone any bright ideas about this? If the gateway goes down, the Hosts will not be able to ping it but they will still be visible to each other on our internal network, will HA ignore the gateway in this instance and not put any hosts into isolation?
