none
S2D VHD Operational Status: Needs Rebalance RRS feed

  • Question

  • Hi Everyone,

    I have a 3 node S2D system with a tiered Storage pool.  All was well until doing a cluster aware update.  The nodes went down one at a time and disk rebuilt after each, but now the disk shows an operational status of "Needs Rebalance" even though the health status is healthy.  I have tried running "Get-StorageSubsystem *cluster* | Get-Volume | Optimize-Volume" which appears to do something (progress bar up top for a bit), but the health status does not change and the disk performance seems to be slower than normal.  

    Looked through this question: https://social.technet.microsoft.com/Forums/Lync/en-US/e08520a8-7994-465c-8646-61c4edaed62e/how-to-rebalance-a-cluster-virtual-disk-in-windows-server-2016?forum=winserverClustering , but I am at the latest CU of RTM.  

    Is there a newer command that gets the virtual disk back on track?

    Tuesday, March 21, 2017 5:42 PM

Answers

  • Posted also on : https://techcommunity.microsoft.com/t5/Windows-Server-for-IT-Pro/S2D-Needs-Rebalance-status/m-p/85008#M270

    and got the following aswer from Elden:

    The fix is now available, go apply this update:
    https://support.microsoft.com/en-us/help/4022723/windows-10-update-kb4022723
    
    Thanks!
    Elden
    


    Andre van den Berg

    • Marked as answer by Thildemar Thursday, July 6, 2017 1:50 PM
    Thursday, July 6, 2017 6:36 AM
  • You don't need to rebuild.  If the virtual and physical disks are healthy and optimize has run, then you are good.  What's happened is that the slabs are healthy and evenly distributed across fault domains (in this case the nodes).  But there is unbalanced number of extents across the physical disks within a fault domain... which really isn't an issue.  Optimize distributes across fault domains, but not within the fault domain.  This is something we are looking to improve forward looking.

    Thanks!
    Elden

    • Proposed as answer by Russel H Monday, June 19, 2017 5:24 PM
    • Marked as answer by Thildemar Tuesday, June 20, 2017 12:08 PM
    Sunday, March 26, 2017 3:14 PM
    Owner

All replies

  • Try running Optimize-StoragePool cmdlet and then you can monitor progress with the Get-StorageJob cmdlet.

    Thanks!
    Elden

    Wednesday, March 22, 2017 2:56 AM
    Owner
  • Did that.  Rebalance runs matching the total volume size, no errors.  Failover and clustering manager still reads "Needs Rebalance"
    Wednesday, March 22, 2017 1:59 PM
  • What do you see with Get-PhysicalDisk cmdlet and Get-VirtualDisk cmdlet for the HealthStatus and OperationalStatus fields??
    Friday, March 24, 2017 4:05 AM
    Owner
  • All physical disk show healthy and OperationalStatus of "ok".  Virtual disk shows Healthy, but OperationalStatus Suboptimal.

    At this point I think I am going to bite the bullet and rebuild the disk this weekend.  New collective wisdom in the last few months makes me think this disk should be split up anyway and I may add some NVMe cache while I am at it.

    Friday, March 24, 2017 3:13 PM
  • You don't need to rebuild.  If the virtual and physical disks are healthy and optimize has run, then you are good.  What's happened is that the slabs are healthy and evenly distributed across fault domains (in this case the nodes).  But there is unbalanced number of extents across the physical disks within a fault domain... which really isn't an issue.  Optimize distributes across fault domains, but not within the fault domain.  This is something we are looking to improve forward looking.

    Thanks!
    Elden

    • Proposed as answer by Russel H Monday, June 19, 2017 5:24 PM
    • Marked as answer by Thildemar Tuesday, June 20, 2017 12:08 PM
    Sunday, March 26, 2017 3:14 PM
    Owner
  • Hi Elden,

    So are you basically saying that a constant "Needs Rebalance" message is normal behavior in this case? (So long as optimize-storagepool runs fine otherwise)

    I did end up recreating disks over the weekend as I was able to add some caching drives and split the large vdisk into individual disks per VM so they can be better managed across nodes (and hopefully limit "faults" such as this one to a single vdisk).  So far the new vdisks seem fine.

    Monday, March 27, 2017 3:25 PM
  • Elden,

    Thanks for the useful info. I proposed this as answer.

    Cheers!

    Monday, June 19, 2017 5:25 PM
  • We have the same issue on our S2D cluster after a patch weekend in march i belive it was. Spoke to Cosmos about this and it's a known issue as Elden says. Everything is fine.

    Microsoft will fix this visual bug. Have not recived an update yet on when. June patches did not solve issue.

    Regards
    Jan-Tore

    jtpedersen.com

    Tuesday, July 4, 2017 7:13 PM
  • It's also available in the July patch. And it works :)

    Solves also another issue with replacing a failed drive that does not want to mount in storage pool.

    JT

    Wednesday, July 19, 2017 9:56 AM