How to prevent VM failover of a SQL Clustered server, in a Clustered host enviroment

[tklose]

I have 4 Vsphere 4 enterprise hosts in a VM cluster.

I have a Active/Active Windows SQL cluster server (MSCS) accross 2 of my hosts.

In the event of a HOST failure, I do not want the other HOST to start up a specific VM guest.

I just want the VM to shut down and stay OFF.

Why?

In a MSCS cluster, some of the configuration is on the local host and it is using RAW luns (which are masked out on 2 hosts) .

I do not know what would occur if HA tried to move the VM in the event the HOST stops responding ( and don't wish to find out the hard way)

Because I am already protected in a VM guest cluster enviroment, I don't need failover/restart to occur.

Under the Cluster HA settings:

The 3 HA VM options for Power off VM, Leave VM powerd on, and Shutdown VM all move the VM and try to restart it.

When VM Restart Priorty is set to Disabled, does it prevent moving the host as well as preventing restart?

Please explain.

[weinstein5]

With Restart Priority set to disabled will prevent that VM from restarting in the event of an HA event- also if the VM is using a LUN that is masked from the other hosts in the HA cluster it will also not start because the host it starts must see all LUNs used by the VM whether it is an RDM or not - also if your are using DRS it is good idea to disable DRS for the VMs in this cluster - because DRS might want to vmotion the VM but will not be able to since some of the LUNs used are masked -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[tklose]

Thanks for the quick reply.

I understand that the VM will not restart with it disabled .

However, I need to disable any attempt to MOVE it (either HA or DRS).

I don't want it to try and fail, it could cause problems.

I don't have big enough of an enviroment to move the VM to a host not in the VMCluster.

Can this be done?

[AlanSSSA]

We've played with this before. If you only have 2 hosts you can set Anti-affinity so they will never be on the same host. We ended up putting them on a cluster with no HA, DRS and no VMotion enabled port groups.

[weinstein5]

HA will only try start a VM in the event of an ESX host or if the VM fails - I am assuming you do not have VM HA monitoring enabled and by setting the setting the VM to disable will prevent it from restarting so the VM will always be on the same host unless someone manually moves it -

DRS has something similar in that you can disable DRS for that one VM - I believe it is the per VM Automation - this will prevent DRS from vmotioning the vm to another host in the cluster.

All of these become somewhat moot if I understand your configuration correctly in that you have masked LUNs used by the VM so that only the ESX host that is running the virtual machine can access those LUNs and not any other nodes in the cluster- this will prevent the VM from starting up on any other node in the cluster and manually vmotion the vm to another host

Hope this helps -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[tklose]

Correct, I Have HA Monitoring off. DRS is not enabled yet, but maybe later.

I have 4 hosts in the cluster, 2 of those have the MSCS VM's (one SQL cluster node on each host)

I need to clarify one critical point.

Does the HA attempt to move the VM? (because the C:\ Drive boot partion and config is in the host server vmfs partition)

I don't want HA to attempt moving it and fail, at all because who knows what harm can result.

[weinstein5]

HA does not move the VM it simply attempts to restart the VM on another node in the cluster and for this to work all virtual disks used by the VM must be accessible by all ESX hosts in the cluster - so with boot virtual disk being stored locally on the ESX server HA will not work and for that matter neither will DRS - that is why you should disable DRS and HA for these VMs -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[tklose]

Thats how I hoped it would work...but the documenation was not very clear for this situation.

It is currently set to: VM Restart Priority "Disabled" and Host Isolation Response "Power off"

I think I'm good.

thanks