VM starts up after ESX Host left isolation mode ?

[TonyJK]

When the ESX host left isolation mode, we observe that the VMs are powered up by itself.

We would like to know is it a normal behaviour OR can we change it ?

Thanks

[apatel1]

Hi TonyJK, if you are talking about the isolation response for an HA cluster, this can be set on a per-VM basis in in the cluster settings --> HA section. The default isolation response depends on the version of ESX you're using (I believe it changed to "leave powered on" from "power off" in all versions from 3.5 U2 onward). This action is only taken when a host determines that it is isolated. A host considers itself isolated when it can no longer reach its isolation address, which is, by default, the Service Console default gateway. This per-host setting can be modified in Advanced Settings. Hope that helps!

Please help me out by awarding points for a "helpful" or "correct" response if you feel that it was useful!

-Amit

[TonyJK]

Thank you for your prompt reply.

We are using ESX 3.5 Update 3. We find that the cluster setting is "Turn Off VM" for isolation mode. I believe that it inherits the settings when we upgrade it from ESX 3.0. For critical production VMs, we have already gotten "leave powered on".

We just would like to know can we change the behaviour for those VMs being shut down by ESX Host entering Isolation mode ? Can we prevent it from starting up when the ESX Host (Service Console) resumes network connection ?

Thanks

[apatel1]

Yes, you're correct. The change of the default setting affects clean installations (or new clusters) only. Regarding what happens when the host re-gains connectivity, I believe the idea is that the VMs are assumed to be powered on on other hosts already, and DRS will just do its magic and shuffle things around based on the newly added capacity. Are you sure that the VMs in question aren't just still powered on from before the host was isolated rather than being re-powered on?

Please help me out by awarding points for a "helpful" or "correct" response if you feel that it was useful!

-Amit

[TonyJK]

Yes, we do encounter a problem that one of the VMs cannot start up when the ESX host regain network connectivity. We have to start it up manually.

Thanks

[apatel1]

I think that it's working as designed. I always prefer to work in terms of examples.....

Let's say you have an HA cluster with ESX1, ESX2 and ESX3 as member hosts. Then, let's say ESX1 hosts VM1 and VM2 on it, ESX2 hosts VM3 and VM4 on it, and ESX3 hosts VM5 and VM6 on it. If ESX1 becomes isolated, let's say that the isolation response on both VM1 and VM2 is set to "power off". Those VMs are powered off, and once ESX2 and ESX3 realize that ESX1 is no longer alive, they will attempt to power on VM1 and VM2 according to their resource availability. If ESX2 and ESX3 only have enough resources (RAM and CPU) to power on VM1, VM2 will not get powered on automatically, even if ESX1 eventually comes back (HA uses priority and then alphabetical order to determine the order of VM restoration). This might be what's happening to your VM.

Please help me out by awarding points for a "helpful" or "correct" response if you feel that it was useful!

-Amit

[TonyJK]

Power failure happened last night for both (2 together) in a VI Cluster last night. After power is resumed, VMs on ESX1 started up properly. However, a specific VM (and the only one) on ESX2 doesn't start up. We have to start it up manually.

Do you have any suggestion ?

Thanks