Sign in
Microsoft.com
 
Windows Server TechCenter
 
 
 
Windows Server TechCenter > Windows Server Forums > Hyper-V > Issue with pass though-disks in Server 2008 R2 Hyper-V cluster
Ask a questionAsk a question
Search Forums:
 

AnswerIssue with pass though-disks in Server 2008 R2 Hyper-V cluster

  • Thursday, October 29, 2009 9:06 PMDJL Users MedalsUsers MedalsUsers MedalsUsers MedalsUsers Medals
     
    Sign In to Vote
    0
    Sign In to Vote
    I've just built our production Windows 2008 R2 Hyper-V cluster.  It consists of 4x Dell PowerEdge R610's and an Equallogic iSCSI SAN.  The servers are connected to the SAN using the Broadcom LOM's, which are configured as iSCSI HBA's.  I've installed MPIO and the Equallogic DSM.

    So far I've built four virtual machines, all of which have their system disks stores on the same CSV.  Two of the VMs also have additional LUNs assigned using pass-through disks.

    Both the virtual machines using pass-through disks are logging the following error in the event log:

    Event ID: 1
    Source: VDS Basic Provider
    Error: Unexpected failure. Error code: 490@01010004

    I'm also unable to use the File Services\Share and Storage Management snap-in.  I receive the following error message: "Connection to the Virtual Disk Service failed.  A VDS (Virtual Disk Service) error occurred while performing the requested operation."

    More worryingly, I rebooted one of the virtual machines this evening and it failed to restart.  Cluster Manager would state the VM was on-line/started and you could initially see the 2k8 loading bars but i'd then get a blank black screen.  After a few minutes all the pass through disk resources were marked as failed in cluster manager and the other vm using a pass-through disk fell over.  I then brought all the failed disks back online (right click | bring online) and both virtual machines started quite happily.  The VMs running solely on the CSV were not affected

    I'm also getting a large number of the following 'Information' events logged;
    Event ID: 2
    Source: MPIO
    Added device to \Device\MPIODisk0. DumpData contains the current number of paths.

    Any help would be greatly appreciated.

     

Answers

  • Wednesday, November 04, 2009 7:26 PMDJL Users MedalsUsers MedalsUsers MedalsUsers MedalsUsers Medals
     Answer
    Sign In to Vote
    0
    Sign In to Vote
    Hi - thanks for your responses and sorry for the delay in responding.

    I believe the problem was caused by packet loss on our iSCSI network.  When I trunked the two 10Gbps links between iSCSI switches flow control was disabled on the ports - after re-enabling flow control i'm not seeing any more packet loss and the cluster has been stable since.


    Joseph, the default policy in Windows Server 2008 R2 is for disks to stay offline, and in server core when you bring them online using diskpart they have the read-only attribute set.

    Vincent Hu,  I'm not sure about your comments regarding pass-through disks.  They are pretty vital for large volumes and situations where high performance is needed - ie SQL Server, Exchange and file servers.  All my VM's use CSV volumes for their OS/Boot disk, but have additional pass-though disks for data.  The only downside is that during live migration the machine has to be paused momentarily while the pass-though disk is brought online on the destination node - in my experience this takes less than 1s and I loose about 1 ping if that.

    Regards,
    Daniel

      
      
     

All Replies

  • Thursday, October 29, 2009 9:56 PMJoseph Noga Users MedalsUsers MedalsUsers MedalsUsers MedalsUsers Medals
     
    Sign In to Vote
    0
    Sign In to Vote

    There was a process that you needed to run on the host to make sure the disks were not auto enabled after reboot. I remember reading that somewhere on MSDN.  Let me look and I will post the update.

    This might be what is causing the problem the host is trying to bring the disks online when then need to be offline.

    http://social.technet.microsoft.com/Forums/en-US/winserverfiles/thread/263f3301-4504-4cd0-8469-68fbb788f129/

    http://technet.microsoft.com/en-us/library/cc753703(WS.10).aspx
    Joseph Noga MCITP, MCSE, MCSA, MCTS CCNA,CCDA,CCVP Managing Consultant VSD Technologies
     
  • Friday, October 30, 2009 3:34 AMVincent HuMSFT, ModeratorUsers MedalsUsers MedalsUsers MedalsUsers MedalsUsers Medals
     
    Sign In to Vote
    0
    Sign In to Vote

    Hi,

     

    I confirmed this with our production group, this scenario is supported. I am sorry that I have a misunderstand about the description in the following blog. 


    Cluster Shared Volumes (CSV): Disk Ownership

    http://blogs.msdn.com/clustering/archive/2009/03/02/9453288.aspx


     

    Best Regards,

    Vincent Hu

     

     
  • Tuesday, November 03, 2009 9:41 AMVincent HuMSFT, ModeratorUsers MedalsUsers MedalsUsers MedalsUsers MedalsUsers Medals
     
    Sign In to Vote
    0
    Sign In to Vote

    ·         Hi,

     

    I want to see if the information provided was helpful. Your feedback is very useful for the further research. Please feel free to let me know if you have addition questions.

     

     

    Best regards,

    Vincent Hu

     
  • Wednesday, November 04, 2009 7:26 PMDJL Users MedalsUsers MedalsUsers MedalsUsers MedalsUsers Medals
     Answer
    Sign In to Vote
    0
    Sign In to Vote
    Hi - thanks for your responses and sorry for the delay in responding.

    I believe the problem was caused by packet loss on our iSCSI network.  When I trunked the two 10Gbps links between iSCSI switches flow control was disabled on the ports - after re-enabling flow control i'm not seeing any more packet loss and the cluster has been stable since.


    Joseph, the default policy in Windows Server 2008 R2 is for disks to stay offline, and in server core when you bring them online using diskpart they have the read-only attribute set.

    Vincent Hu,  I'm not sure about your comments regarding pass-through disks.  They are pretty vital for large volumes and situations where high performance is needed - ie SQL Server, Exchange and file servers.  All my VM's use CSV volumes for their OS/Boot disk, but have additional pass-though disks for data.  The only downside is that during live migration the machine has to be paused momentarily while the pass-though disk is brought online on the destination node - in my experience this takes less than 1s and I loose about 1 ping if that.

    Regards,
    Daniel