Hi,

I have a serve Problem with SCVMM 2012 R2 UR5. We have two node Hyper-V cluster with JOBOD and Storge Spaces as VM storage. (Two server connected with SAS to one JBOD, clustered storage spaces as CSV, VMs storage on the CSV)

Since UR5 we a problem with Live Migration through SCVMM. Here is an example:

• VM is running on host 1 and stored on CSV1
• Host 1 is the owner node of CSV1

  Live Migration from host 1 to host 2 works. Live Migration from host 2 to host 1 fails with Error 2404 (VMM cannot find a required object.)

The root of this seems to be the CSV on storage spaces. Because the when the owner of the CSV changes the whole Disk (storage spaces) moves from one host to the other. This detects SCVMM and for the host which is not the owner of the CSV SCVMM has a zero value for the HostDiskID of the CSV in his database.

Has anyone a similar configuration and the same problems?

Thx

J0F3

Hello,

Live migration works with the Failover Cluster Console or you have the same problem?

UR5 is a VMM update, and not a Windows or Hyper-V update, so there is 2 options so far :

- The UR5 did something wrong with VMM that led him to fail live migrations

- Coincidentally, ther's a problem with your failover cluster

Try to live migrate your VM using the Failover Cluster console, and only one possibility will remain

Yes Live Migration in Failover Cluster Manager still works perfect. So it's definity a issue in SCVMM since UR5. With UR4 we didn't had that problem.

In the mean time I could find exactly what's going wrong in SCVMM. I could it narrow down to the HostDiskID column in the tbl_ADHC_HostVolume table in the SCVMM DB. If the Hyper-V host is not the owner of the CSV the value HostDiskID in the DB is "00000000-0000-0000-0000-000000000000". Naturally SCVMM can then not find any disk with the ID "00000000-0000-0000-0000-000000000000" in the DB and that's why the job in SCVMM fails with the error 2404 (VMM cannot find a required object.).

Please see the following hotfix that should address these know issues with UR5 for VMM:

https://support.microsoft.com/kb/3039296?wa=wsignin1.0

-kn

I'm seeing the same issue, after installing UR5. Could you let post if you come up with a resolution, please?

Do you encounter the issue even when live migrating from the failover cluster console ?

This means, when you initiatea Live Migration from the VMM console, the migraion fails.

What is the error message in the VMM events ?

What is the storage you are using ?

Sure, I will post it here if I have solution. Do you also use storage spaces as VM storage?

Reading over your post again the configuration and symptoms seem near perfect matches. However, we're using an iSCSI SAN. The error message is the same as above.

Hello!

We have same problems, and not only with cluster live migration, but with Live VSM migration too.

So, here what I've discovered.

Suppose we have HostA and HostB. HostB has disk "\\?\Volume{GUID}" , "C:\" (on one physical disk) and "D:\" (on another physical disk) in Hardware tab in host properties. Disks are simple local RAID disks.

But when we go to Storage tab - we only see the first physical disk with two volumes ("C:\" and "system reserved"). As already mentioned, HostDiskID for volume "D:|" is zero-guid in VMM database.

When migrating VM from hostA to hostB it is possible to select "d:\" as a target location, but migration fails just after start with error:

We tried remove/add host in VMM, remove/add disk on host, reformatting disk, but only solved this problem by recreating RAID volume.

Today I've noiced migration problems in cluster. I can confirm, that live migraion works fine in Failover Cluster Manager. Also, only migration to the host that owns CSV on which VM is located works in VMM.

• Edited by Demonixed 20 hours 57 minutes ago

• Edited by J0fe 19 hours 33 minutes ago

Hello!

We have same problems, and not only with cluster live migration, but with Live VSM migration too.

So, here what I've discovered.

Suppose we have HostA and HostB. HostB has disk "\\?\Volume{GUID}" , "C:\" (on one physical disk) and "D:\" (on another physical disk) in Hardware tab in host properties. Disks are simple local RAID disks.

But when we go to Storage tab - we only see the first physical disk with two volumes ("C:\" and "system reserved"). As already mentioned, HostDiskID for volume "D:|" is zero-guid in VMM database.

When migrating VM from hostA to hostB it is possible to select "d:\" as a target location, but migration fails just after start with error:

We tried remove/add host in VMM, remove/add disk on host, reformatting disk, but only solved this problem by recreating RAID volume.

Today I've noiced migration problems in cluster. I can confirm, that live migraion works fine in Failover Cluster Manager. Also, only migration to the host that owns CSV on which VM is located works in VMM.

• Edited by Demonixed Wednesday, March 11, 2015 10:12 AM

Hello!

We have same problems, and not only with cluster live migration, but with Live VSM migration too.

So, here what I've discovered.

Suppose we have HostA and HostB. HostB has disk "\\?\Volume{GUID}" , "C:\" (on one physical disk) and "D:\" (on another physical disk) in Hardware tab in host properties. Disks are simple local RAID disks.

But when we go to Storage tab - we only see the first physical disk with two volumes ("C:\" and "system reserved"). As already mentioned, HostDiskID for volume "D:|" is zero-guid in VMM database.

When migrating VM from hostA to hostB it is possible to select "d:\" as a target location, but migration fails just after start with error:

We tried remove/add host in VMM, remove/add disk on host, reformatting disk, but only solved this problem by recreating RAID volume.

Today I've noiced migration problems in cluster. I can confirm, that live migraion works fine in Failover Cluster Manager. Also, only migration to the host that owns CSV on which VM is located works in VMM.

• Edited by Demonixed Wednesday, March 11, 2015 10:12 AM

Hello!

We have same problems, and not only with cluster live migration, but with Live VSM migration too.

So, here what I've discovered.

Suppose we have HostA and HostB. HostB has disk "\\?\Volume{GUID}" , "C:\" (on one physical disk) and "D:\" (on another physical disk) in Hardware tab in host properties. Disks are simple local RAID disks.

But when we go to Storage tab - we only see the first physical disk with two volumes ("C:\" and "system reserved"). As already mentioned, HostDiskID for volume "D:|" is zero-guid in VMM database.

When migrating VM from hostA to hostB it is possible to select "d:\" as a target location, but migration fails just after start with error:

We tried remove/add host in VMM, remove/add disk on host, reformatting disk, but only solved this problem by recreating RAID volume.

Today I've noiced migration problems in cluster. I can confirm, that live migraion works fine in Failover Cluster Manager. Also, only migration to the host that owns CSV on which VM is located works in VMM.

• Edited by Demonixed Wednesday, March 11, 2015 10:12 AM

• Edited by J0fe Wednesday, March 11, 2015 11:37 AM

• Edited by J0fe Wednesday, March 11, 2015 11:37 AM

• Edited by J0fe Wednesday, March 11, 2015 11:37 AM

Are your hosts using basic or dynamic disks?

After applying CU5, we could no longer live migrate to hosts that had dynamic disks for storage. After reformatting as basic disks, we were able to live transfer again.

Our hosts have basic disk formatted as CSV (CSVFS). 

• Marked as answer by J0fe 23 hours 6 minutes ago

• Marked as answer by J0fe Monday, March 16, 2015 8:04 AM

• Marked as answer by J0fe Monday, March 16, 2015 8:04 AM

• Marked as answer by J0fe Monday, March 16, 2015 8:04 AM