none
Unable to correct cluster quorum settings RRS feed

  • Question

  • I have a Windows Server 2012 cluster with 2 servers, a witness disk, and a storage disk. The quorum is set to Node majority but the recommended setting is Node and Disk majority. This can't be set however because the wizard says "an appropriate disk could not be found".

    How do I correct this?


    Jonathan

    Thursday, December 26, 2013 10:34 PM

All replies

  • What type of shared storage are you using for the witness and storage disks?  Are both disks properly configured to be shared between both nodes of the cluster?  How do you have the shared storage configured?  What sort of warnings/errors did you get when you ran the cluster validation wizard before creating the cluster?

    .:|:.:|:. tim

    Friday, December 27, 2013 4:35 PM
  • What type of shared storage are you using for the witness and storage disks?  Are both disks properly configured to be shared between both nodes of the cluster?  How do you have the shared storage configured?  What sort of warnings/errors did you get when you ran the cluster validation wizard before creating the cluster?

    .:|:.:|:. tim

    I'm not certain what you're looking for when you ask "what type" of shared storage or how the shared storage is configured. The 2 disks are available to both nodes of the cluster and this cluster has been in production for about 9 months.

    And I'm sorry, but in no way can I remember what the results of the wizard were before creating the cluster. In fact I'm not even sure the wizard was run at that point. It was the first cluster we have ever created and if the MS instructions we got from the web did not include running the wizard, then we probably didn't run it. But even if we did, that is definitely too far back to recall something like that. Again, sorry.


    Jonathan


    Friday, December 27, 2013 5:02 PM
  • In fact I'm not even sure the wizard was run at that point. It was the first cluster we have ever created and if the MS instructions we got from the web did not include running the wizard,"

    If the instructions did not include instructions to run the validation wizard, it most likely was not Microsoft documentation.  For a cluster to be supported, it must pass the validation wizard.  If you make a support call to Microsoft, one of the first things they will ask to see is the output from the validation wizard.  You should always have the last two or three runnings of the validation wizard available.  A new validation wizard should always be run before and after making changes to a running cluster.  The wizard performs hundreds of tests and will generally find configuration errors.

    All is not lost.  You can run the validation wizard even on a running cluster.  And, because you are only interested in the Witness disk, you can select just the witness disk when it asks which disks to test.  This should not impact the resource that is using the data disk, unless the witness disk is improperly set up to be used by the resource.

    "I'm not certain what you're looking for when you ask "what type" of shared storage or how the shared storage is configured."

    Is the storage an iSCSI SAN, Fibre Channel, Fibre Channel over Ethernet, SAS, PCI RAID, or SMB3?  Those are the types of shared storage that a Failover Cluster can use.  It is interesting that you say it has been working for 9 months.  Have you ever failed over the resources from one node to the other?  What are the resources you are clustering?  What does the Failover Cluster manager show when you select Storage?  Can you see two volumes in storage on both nodes of the cluster in Disk Manager?  You will not be able to access them from both nodes, but you should be able to see both of them.  Only one of them should be able to access a disk at a time.  This could be one accessing one disk and the other accessing the other disk or one node accessing both disks.


    .:|:.:|:. tim


    Saturday, December 28, 2013 4:11 PM
  • If Microsoft instructions say to run the wizard then it was run as the instructions were slowly and carefully followed since this was the first time we set one up.

    Thanks for the tip about running the wizard. I've run it multiple times but the directions say the VMs storage volume have to be shut down. But that causes problems on its own since that shuts down both DCs. I will run the wizard again selecting only the witness disk. The witness disk is setup ONLY to be a witness disk. The resources (VMs) are all set up to use the storage volume.

    Yes we have failed the VMs back and forth many times. The VMs are all servers mostly running Windows Server 2012 with one running 2008 R2 and 2 running 2003. Storage in the FCM show Disks and Pools. Disks shows Cluster Disk 1 and Cluster Disk 2, 1 is the Witness disk and 2 is the Storage disk. There are no pools. In Disk Management I can see both Disks even though one of the nodes is the owner of both disks. The status of the disks is Reserved in DM on both nodes and the disk icon has a red arrow pointing down for the disks on the node that is NOT the owner. This is to be expected.

    I hope this answers your questions and helps to figure out why the quorum wizard says a disk can't be found.


    Jonathan

    Sunday, December 29, 2013 12:49 AM
  • "but the directions say the VMs storage volume have to be shut down. But that causes problems on its own since that shuts down both DCs."

    I know that Microsoft made it possible to store DCs on clusters, but I don't recommend it.  I have seen issues with making DCs as highly available cluster resources.  Are a result, though I only use DCs as VMs in my environment, I never make them highly available.  There is really no need to make them highly available as they have their own built-in HA capability.  I haven't tried with 2012 R2, but I had too many problems with trying to make DCs as HA resources in a cluster that I simply stopped doing it.

    As for the directions saying that the VM volume needs to be shut down, that is optional.  During the running of the wizard, it will display the disks available to the cluster.  In your case, it appears that both disks are showing in the cluster manager.  You have the option to select which of the disks to use in the tests.  Simply select the witness disk and not the data disk and the validation wizard will run fine without even touching the data disk.

    Did you happen to make both of the disks as Cluster Shared Volumes?  If so, the witness disk should NOT be configuring that way.  That could be a cause of the problem.  If it is configured as a CSV, remove it.  It will remain as a cluster disk, but simply not as a CSV.  It would then be available to chose as a witness disk.


    .:|:.:|:. tim

    Monday, December 30, 2013 2:29 PM