Hyper-V - Persistent Reservations & Corrupt VHDs

  • 31 January 2017
  • 5 replies
  • 1262 views

Badge +5
How do Persistent Reservations work with Hyper-V on Nutanix?
  • I'm asking because with our other "Hyper-V with SMB Storage" clusters, you are not able to use Hyper-V manager on HOST1 to modify VHDs that belong to a VM running on HOST2, as you would expect - Error reads: "Access Denied". However, with Nutanix, we are able to do this and it leads to VHD corrouption upon next reboot of the VM
  • Here is the scenario:[list]
  • Hosts: 5 node Nutanix cluster with Hyper-V (Dell XC hardware)
  • VM1 is running on HOST1
  • Using HOST2, launch Hyper-V manager (not Failover Cluster Manager). Edit Disk to resize a VHD of VM1. Resize successful (This fails in our non-Nutanix clusters)
  • Using HOST2, check the properties of VM1’s VHD and you will notice that the file size has changed
  • Using HOST1 (where the VM is running), check the properties of the VHD and you will notice that the file size HAS NOT changed
  • Reboot VM1, server will not boot – VHD is now corrupt
[/list]Is this a known issue/bug?
I understand that you should never use Hyper-V Manager to modify clustered VMs but the fact that you can (whereas our other clusters don't allow it) is conscerning.

-E

This topic has been closed for comments

5 replies

Userlevel 3
Badge +17
Hello,

We are aware of this behavior and are developing a solution to be released soon. While the Nutanix SMB share is not intended for external access we are implementing the required SMB locking mechanisms to ensure your scenario is not possible. Please contact support if you need more details.

Thanks,
Mike
Badge +5
Thank you - do you know if there is a bug number or KB article on this issue?
Userlevel 3
Badge +17
Yes, with support you can reference ENG-23074.

Thanks,
Mike
Userlevel 6
Badge +29
Quick update, we're getting close on this one. Literally have code getting committed today. Will have to go through the standard QA cycles, but its looking good thusfar.
Badge +5
That's great news - thanks for the update!