Announcement

Collapse
No announcement yet.

Cluster Shared Volumes - Event 5125 with Chinese characters?

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

  • Cluster Shared Volumes - Event 5125 with Chinese characters?

    Hi,

    I've never set up CSV before however assumed it would actually work. In other setups we have Xen Server accessing shared storage on a SAN.

    In this setup we have three physical machines and a 10Gb iSCSI SAN.

    Each physical machine has two direct 10Gb connections to the SAN on a unique subnet (no top of rack switch). Multipathing is enabled and all three servers see both routes to the SAN. That much is good.

    iSCSI initiator is all configured up and I can see the LUNs. I can bring each LUN online on any one of the three servers and use it then take it offline and bring it up on any of the other servers to confirm that it all works.

    The problems started when I started using FCM. The cluster created fine and I was able to add the disks then add them to CSV. Since doing that FCM has been completely unstable.

    Problem 1) I tried to take the disks offline so that I can make sure the labelling is correct from the SAN RAID through to the partitions and mount points. Three of the disks went offline. The others went Offline Pending and I had to leave them overnight. They were offline come morning. I went through re-labelling and brought them back online to find out those three volumes now showed as RAW filesystems. Can't have that happening to production filesystems!

    Problem 2) After leaving everything online for a while I checked back on the servers to find most of the LUNs in Online (Redirected). Pretty bad for performance.

    Problem 3) I just went back in to FCM to have it up as reference whilst typing this and all three nodes were offline with the cluster service stopped. The events logged for this are 1177 that the quorum was lost. I set up a file server witness on a physical server that is not involved in the cluster. This server also has a separate witness for the Exchange DAG and has not had any issues.

    Problem 4) Even when they are online, I have only managed to actually change into the mount point of a CSV once. Currently my test disk is Online (No Access) and the mountpoint is visible on the owner node. Clicking on it acts as though it isn't there and the mountpoint is not visible on any other node at all.

    Problem 5) I get a lot of event id 5125 complaining about active filter drivers. Unfortunately the only thing listed in "Active filter drivers found" is a load of chinese symbols.

    I've read through numerous pages and guides from MS and elsewhere and all indicate that I've done things properly. I think all the issues with things being inaccessible and going offline can be tied to the event id 5125 but with it displaying Chinese characters I don't really know where to start. The servers are all Server 2012 R2 and apart from the Hyper-V role they run Vipre antivirus (Managed Antivirus from LogicNow) and StorageCraft ShadowProtect SPX for backups.

    Apart from uninstalling the backup software and the antivirus does anyone have any suggestions?

  • #2
    Your quorum (witness) has to be accessible as a clustered disk, same as the data volumes you're storing everything else on. The best way to do it would be to provision a LUN on your SAN at 1Gig space, and set that as the quorum. If you had had that provisioned when you ran the preparation wizard for setting the cluster up, the wizard would have chosen that volume for the quorum on it's own.
    *RicklesP*
    MSCA (2003/XP), Security+, CCNA

    ** Remember: credit where credit is due, and reputation points as appropriate **

    Comment


    • #3
      Originally posted by RicklesP View Post
      Your quorum (witness) has to be accessible as a clustered disk, same as the data volumes you're storing everything else on. The best way to do it would be to provision a LUN on your SAN at 1Gig space, and set that as the quorum. If you had had that provisioned when you ran the preparation wizard for setting the cluster up, the wizard would have chosen that volume for the quorum on it's own.
      I don't think that's correct. The Cluster Quorum Wizard has an option to use a file share witness for the Quorum.

      Comment


      • #4
        With the errors reported, it sounds like access to the witness is being interrupted. So he's either got network or workload issues with access to the Witness, or the CSV LUN. I suggested the Witness location because A) it's gonna be 10Gig connectivity with his hardware and B) his current Witness is on a server which may have other issues. Remove the witness from that non-ideal setup seemed the easiest first step.
        *RicklesP*
        MSCA (2003/XP), Security+, CCNA

        ** Remember: credit where credit is due, and reputation points as appropriate **

        Comment


        • #5
          Originally posted by RicklesP View Post
          With the errors reported, it sounds like access to the witness is being interrupted. So he's either got network or workload issues with access to the Witness, or the CSV LUN. I suggested the Witness location because A) it's gonna be 10Gig connectivity with his hardware and B) his current Witness is on a server which may have other issues. Remove the witness from that non-ideal setup seemed the easiest first step.
          My bad, I thought you were suggesting that FC could only use a disk witness and not a file share witness.

          Comment


          • #6
            I realised I never came back here.

            Re the witness, I did try with both file server witness and a lun on the SAN before posting.

            The issue is with StorageCraft ShadowProtect. The software was installed for taking DR images of the hypervisors. Even with backups disabled it appears the drivers that this software installs interferes with CSVs.

            Uninstall ShadowProtect and it works fine. I have a ticket open with StorageCraft regarding it, the last I heard they were replicating the setup in the lab to investigate.

            Comment


            • #7
              Thanks for the follow-up.
              *RicklesP*
              MSCA (2003/XP), Security+, CCNA

              ** Remember: credit where credit is due, and reputation points as appropriate **

              Comment

              Working...
              X