Announcement

Collapse
No announcement yet.

Failing Over a VM on a Geoclustered Hyper-V Host

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Failing Over a VM on a Geoclustered Hyper-V Host

    I've racked a 2008 server in each of our 2 data centers, and setup a majority node set cluster with file witness share using a share on our HQ DC. This negates the need to have common storage across geographical locations.

    What's not entirely clear from the Hyper-V documentation is whether I need to setup a VM on each of the servers in order to enable failover for the VM.

    I'm thinking not, as the second node is passive, but how does it know to start the VM (and where does it get the VMC/VHD files) if the active node fails?

    Do I need to use VSS to make continuous copies of the VMC/VHD files from the active node to the passive node?

    Any help clarifying this would be welcome.

  • #2
    Re: Failing Over a VM on a Geoclustered Hyper-V Host

    Originally posted by TokyoBrit View Post
    I've racked a 2008 server in each of our 2 data centers, and setup a majority node set cluster with file witness share using a share on our HQ DC. This negates the need to have common storage across geographical locations.

    What's not entirely clear from the Hyper-V documentation is whether I need to setup a VM on each of the servers in order to enable failover for the VM.

    I'm thinking not, as the second node is passive, but how does it know to start the VM (and where does it get the VMC/VHD files) if the active node fails?

    Do I need to use VSS to make continuous copies of the VMC/VHD files from the active node to the passive node?

    Any help clarifying this would be welcome.
    With Failover Clustering, AFAIK, there needs to be a shared storage between the 2 nodes that appears in Disk Management as disks. Your VMs are placed there. Should the node the VMs are running on fail, his will be detected in a matter of seconds and other failover cluster members will take ownership and bring them online automatically.

    Hyper-V gives you the option of creating shadow copies and is certainly a worthwhile procedure to carry out, especially before carrying out any changes to a VM.

    Failover Clusering can also be used within the VMs themselves, should a service or some other failure occur.

    Comment


    • #3
      Re: Failing Over a VM on a Geoclustered Hyper-V Host

      I thought that Majority Node Set with File Share Witness clustering removed the requirement for having shared storage, or is it that the shared storage could be a network share on another server?

      I'm trying to avoid having to setup an expensive geo-SAN.

      And from your comment, it looks like I could create a cluster using VM's within the host, taking into account I would need further Data Center/Enterprise licenses

      What I can't see is how a VM-cluster would fail a VM-node without some underlying H/W problem that would also affect the other VM-node...

      Comment


      • #4
        Re: Failing Over a VM on a Geoclustered Hyper-V Host

        Your right, failover clustering with majority node set prevents a single point of failure for the quorum hosting the services. This can now be a share on a 3rd party PC and should a quorum fail, other nodes can be set to be permitted to assess that a node is down and initiate the required action to restore services on another node.


        This is a worthwhile read.

        http://www.informit.com/articles/art...45125&seqNum=9

        It describes the process regading majority node. If a host fails, it is detected as such and will then be isolated from participating in the cluster, so won't effect other nodes in the cluster.

        With regards to licensing, you can double check with Micrsoft, but I would have thought you shouldn't need any further licensing. Enterprise includes 4 licenses for VMs and Datacenter edition, limitless. With it being a failover cluster, I would have thought that you would have he same number of CALs etc. The other failover cluster members are only used, if the other node fails, so therefore, won't have he ability to service clients. I know that Exchange 2007 licensing works this way.

        I must admit, I have only used a SAN for failover clustering. Whether you can use something such as doubletake (you mentioined whether VSS can be used to replicate the VMs) to allow VMs to replicate VMs between the nodes, so allowing the passive node to mount the VMs should the active node fail, I wouldn't know. VSS can't be used in that way as a service used for taking local snapshots only.

        Comment


        • #5
          Re: Failing Over a VM on a Geoclustered Hyper-V Host

          Thanks. I took a look at that article, then did some googling on good book for Hyper-V and bought a couple.

          I think that 2008 and Hyper-V gives me the flexibility to pretty much do what I need, and would allow us to move some of our 50+ servers off physical hardware and into virtual farms.

          But I just need to ensure high availability for a majority of the services that our servers currently provide.

          Comment


          • #6
            Re: Failing Over a VM on a Geoclustered Hyper-V Host

            Originally posted by TokyoBrit View Post
            Thanks. I took a look at that article, then did some googling on good book for Hyper-V and bought a couple.

            I think that 2008 and Hyper-V gives me the flexibility to pretty much do what I need, and would allow us to move some of our 50+ servers off physical hardware and into virtual farms.

            But I just need to ensure high availability for a majority of the services that our servers currently provide.
            Good luck with the move. I went to a Microsoft Technet Roadshow regarding their virtualisation the other day. It looked very impressive. The speaker stated that the whole of the Technet site was already virtualised including MSDN and half of the Microsoft site. (don't quote me on this though, this was what I was told)

            http://blogs.technet.com/jamesone/ar...materials.aspx

            They also reference some software that can convert a normal, partitioned storage away in to a SAN. Maybe be useful for you.

            BTW, it would be good if you were able to post back as well with how you got on etc. and procedures you carried out. It would be found interesting by forum users.

            Comment

            Working...
            X