We(VMware PSO) had a discussion yesterday on the fact whether it’s supported to have MSCS(Microsoft Clustering Services) VM’s in a HA/DRS cluster with both HA and DRS set to disable. I know many people struggle with this because it doesn’t make sense in a way. In short: No, this is not supported. MSCS VM’s can’t be part of a VMware HA/DRS cluster, even if they are set to disabled.
I guess you would like to have proof:
For ESX 3.5:
http://www.vmware.com/pdf/vi3_35/esx_3/r35u2/vi3_35_25_u2_mscs.pdf
Page 16 – “Clustered virtual machines cannot be part of VMware clusters (DRS or HA).”For vSphere:
http://www.vmware.com/pdf/vsphere4/r40/vsp_40_mscs.pdf
Page 11 – “The following environments and functionality are not supported for MSCS setups with this release of vSphere:
Clustered virtual machines as part of VMware clusters (DRS or HA).”
As you can see certain restrictions apply, make sure to read the above documents for all the details.
Hello Duncan,
Could you provide more information why the solution is not supported and are there plans for changing that policy? I think that sometimes MSCS is necessity and it does not make a lot of sense to have separate environments for clustered virtual machines and for machines that we want to be protected with HA for example.
Best regards
Alex
Sorry that’s NDA info and can’t go in to details exactly why / when etc.
Here is hoping they change this soon. I am going to have to have two tiers of clusters in a data center to account for MCSC and non-MCSC VMs
Duncan, according to the VMware’s “Deploying Exchange Server 2007 on VMware Infrastructure: A VMware Internal Case Study” (http://www.vmware.com/files/pdf/solutions/08Q4_VM_Exchange_Server_2007_VI3_WP.pdf) MSCS
Duncan, according to the VMware’s “Deploying Exchange Server 2007 on VMware Infrastructure: A VMware Internal Case Study” (http://www.vmware.com/files/pdf/solutions/08Q4_VM_Exchange_Server_2007_VI3_WP.pdf) MSCS can be used with HA and DRS.
That seems inconsistent with the documents that you mentioned.
Is there an explanation for that?
Which section exactly? I see a reference to CCR. Anyway if MSCS is being used the VMs can’t be a part of the HA/DRS cluster. (Technically they can but it’s just not supported.)
CCR is used in Exchange 2007 MSCS clusters, so as Mike says, this document seems to conflict with the documents you reference.
“VMware required a rapid response to any failure ranging from a
single resource to loss of an entire site. The legacy environment
relied on a single-site design with a tape-based recovery
process. The combination of VMware HA and the application
failover capabilities built into Exchange Server 2007 CCR can
effectively eliminate single points of failure inherent in a nonvirtualized
implementation, and assure a rapid return of services
in scenarios less dire than a true regional catastrophe.”
It seems ashame to have that limitation as vMotion is great to manage the virtual infrastructure, and VMwareHA is good for Tier 2 and Tier 3 apps, but if you need to ensure your Tier 1 apps are available, you still need MSCS or other application level monitoring to ensure the availability of your applications.
I can undersatand why DRS is not supported in a cluster environment but why not HA.
A machine fails, MSCS cluster kicks in, HA brings the machine up else where.
It just doesn’t really make sense.
Can anyone expand why HA (just HA and not DRS) is not supported by vmware?
Thought I’d chime in since it looks like this post hasn’t been updated and I ran across it studying for my VCAP-DCA.
MSCS nodes in an HA and DRS cluster are now supported with 4.1 thanks to the VM-VM and VM-Host affinity functionality. vMotion is still not recommended.
http://www.vmware.com/pdf/vsphere4/r41/vsp_41_mscs.pdf
“Sorry that’s NDA info and can’t go in to details exactly why / when etc.”
BLAHHHHHH!!!
You just don’t know!
Well I do know that you can run them in a HA cluster today, as long as you set the VMs itself to “disabled” as being part of HA/DRS.
Duncan,
I understand why we would disable DRS due to the stunning period and MSCS may thing the VM is down etc etc… However I fail to understand why HA would need to be disabled. Wouldn’t that just restart the VM on a different host based on the affinity rules, and when it does, what problems could it raise?
hi
my name is ferenc, I am at the very beginning in clustering matter. I would like to find out if is there any way to build an ha cluster from vmware esx or xenserver. if one fails the othe pc can take over the failed one.