locked
Server 2012 Failover Clustering Not Failing Over RRS feed

  • Question

  • Running a simple two-node Server 2012 Datacenter Cluster for Hyper-V Clustering.  I'm able to live migrate the vms back and forth between the clustered nodes.  When I simulate a test failover or an actual failover, the failover cluster does not move the cluster services to the second cluster node, effectively dropping my vms.

    I've checked my Cluster Core Resources and both nodes are checked as possible owners.

    Simulating the failure triggers a "cluster network name is not online" error, then reverts back to the original host.

    Also getting a "Cluster resource "Cluster Disk 1 in Cluster Group failed error event.  This is my Quorum Disk.

    Thoughts, suggestions?  Thanks in advance.


    Fuel

    Monday, September 30, 2013 12:07 AM

Answers

All replies

  • Hi,

    A Hyper-V Cluster works differently. If a host fails the VMs are restarted on the other. And you are using Cluster Shared Volumes, it is no matter which server has the resources. Both can access the CSV simultaneously.

    Monday, September 30, 2013 6:11 AM
  • Running a simple two-node Server 2012 Datacenter Cluster for Hyper-V Clustering.  I'm able to live migrate the vms back and forth between the clustered nodes.  When I simulate a test failover or an actual failover, the failover cluster does not move the cluster services to the second cluster node, effectively dropping my vms.

    I've checked my Cluster Core Resources and both nodes are checked as possible owners.

    Simulating the failure triggers a "cluster network name is not online" error, then reverts back to the original host.

    Also getting a "Cluster resource "Cluster Disk 1 in Cluster Group failed error event.  This is my Quorum Disk.

    Thoughts, suggestions?  Thanks in advance.


    You're a little bit confused... VMs don't expect to spawn on another physical host if their primary physical host (the one they run on now) is down. They will indeed be re-started and Denny is 100% correct on this. So if you need zero downtime you can go basically a couple of routes. First one is guest VM cluster. You run TWO (at least) VMs each bound to the different physical host and you run Windows cluster between them (or use application built-in clustering features, think about Microsoft SQL Server and Database Mirroring thing as an example). Then if physical host will go down all the control would move to another VM running on the other physcal host. No downtime! Also upcoming R2 now can do shared VHDX so you don't need to expose iSCSI or FC volume to your guest VM cluster, everything is very easy now. See:

    http://blogs.technet.com/b/keithmayer/archive/2013/03/21/virtual-machine-guest-clustering-with-windows-server-2012-become-a-virtualization-expert-in-20-days-part-14-of-20.aspx#.Uklv4Ya-2r0

    http://blogs.technet.com/b/mghazai/archive/2009/12/12/hyper-v-guest-clustering-step-by-step-guide.aspx

    http://www.petri.co.il/create-guest-clusters-windows-server-2012-hyper-v.htm

    ...and about R2 shared VHDX thing:

    http://technet.microsoft.com/en-us/library/dn265980.aspx

    http://www.aidanfinn.com/?p=14936

    If for some reason guest VM cluster is not a way to go you'll have to switch a hypervisor... VMware ESXi has Fault Tolerance and that's the thing

    you may really look for. It basically mirrors both VMs on a very small granularity level (including CPU registers stack) so failover happens really fast

    and good news FT does work for ANY virtual machine for ANY application and with NO virtual machine re-configuration or inside-a-VM management. See:

    http://www.vmware.com/products/vsphere/features-fault-tolerance

    http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1013428

    Hope this helped :)


    StarWind iSCSI SAN & NAS

    • Proposed as answer by VR38DETTMVP Monday, September 30, 2013 12:42 PM
    • Marked as answer by Jeremy_Wu Sunday, October 6, 2013 7:15 AM
    Monday, September 30, 2013 12:42 PM
  • Understood.  Good points and the links are very informative.  When I originally setup the physical host servers with the cluster and hyper-v roles, I imported a single VM over to the first node and added the vm as cluster role for initial testing.  I was able to reboot node one and the cluster services, roles, and disks moved over to node two. 

    True to your point, the VMs did reboot but this is not my issue yet, at least until zero downtime becomes my focus.  If I stripped it down to just the clustered share volume (currently seen by both servers) and rebooted node 1, the cluster itself would not fail over to node 2.

    When rebooting Node 1, I can literally see on Node 2 where the entire cluster in Failover Cluster Manager drops and I'm not able to reconnect to it.  What originally failed was the client access network was not configured for clients to access it.  This prevented the cluster from starting.  Fixed this, but the cluster itself does not fail over.


    Fuel

    Monday, September 30, 2013 8:52 PM