Instant VM Recovery hung

[stepkinetic]

I will try to give all the pertinent info without writing a novel:

We had a disaster recovery situation, and we had to bring our file server up in instant recovery due to it's size and the time we had for production to resume. Once things had calmed down, we migrated our VM to a production LUN on our VNX 5500. In vsphere, the VM in question shows it's on "LUN5" and associated files in that datastore have a timestamp that is current (today, 9/3). It is running on host "esx6".

Unfortunately, after appearing to run successfully, we get this in veeam: "VeeamAgent Terminated Unexpectedly" and the VM still showing as mounted in Instant Recovery on host "esx6".

My course of action to resolve this, will be to wait for our maintenance window this Sunday, and unpublish the VM. We have successful backups on the VM in Veeam dated today, and should up until Sunday. If we run into any issues, the plan will be to restore the VM from backup.

Does anyone see any issues with this, or have a better course of action that can be followed to uncover what's going on? I'm just nervous due to the nature of what happens when you unpublish a vm in instant recovery. Thanks all in advance

[stepkinetic]

Let me clarify a little more. When I say "after appearing to run successfully", I am talking about the migration from veeam instant recovery back to production.

[foggy]

Ryan, I'd check with support regarding the error (you're getting it during the IR session, right?) to define the reasons of why does it occur. If you have successful backup of the VM after migration (and have verified their recoverability), I believe you're safe with trying to unpublish it.

[dkrimmer]

Just being curious: how did you solve your issue? And, btw what's wrong with using Storage VMotion to migrate the running VM from the temporarily mounted NFS datastore to the desired production LUN once the VM was up and running? This is (if available) in fact what IR does when selecting "Migrate to Production".

[stepkinetic]

We haven't solved the issue yet, and I'll be removing this from IR this Sunday. We did in fact use storage vmotion to move the VM, and it was successful it appears, judging by where the vm is running from in vsphere. However the IR session for the VM still shows as running in veeam. SO my worry was that once I unpublish it, it'll cause issues with the VM in vsphere.

We do in fact have repeated successful backups so at this point I'm not too worried. Once we complete the unpublish procedure on Sunday, I'll let everyone know how it went. I'll also reach out to veeam support for any additional advice.

[foggy]

We did in fact use storage vmotion to move the VM, and it was successful it appears, judging by where the vm is running from in vsphere. However the IR session for the VM still shows as running in veeam.

Just to check: you've migrated the VM from within vSphere Client, not using Veeam B&R Quick Migration?

[stepkinetic]

That's correct. so I guess this would be the better question:

I have an instant recovery that ran for our production server, the server was vmotioned onto a datastore. Since Veeam migration utility wasnt used, can I stop the session for instant recovery without causing an outage of the server.

[foggy]

Well, since instead of Quick Migration, you've used storage vMotion to migrate VM to production, now you're in situation where Veeam B&R doesn't know that the VM is not running from backup and Instant Recovery session is still not completed. If you simply unpublish the VM, it will be deleted. You can try to migrate the VM again by right-clicking the IR session and selecting Migrate to production to correctly finalize Instant Recovery, however I'm not sure this works this way (better test on some other VM first).

[stepkinetic]

From support:

"The best way to determine if the Instant recovery is still reliant to the running session in Veeam would be opening a vSphere client, edit the VM settings for the instant recovery, and check the VM's disks. If they are still pointing to Veeam's NFS datastore that is created upon starting the IR then "stopping the session" WILL cause additional issues. IF all disks on that IR have been moved over to a production datastore then stopping the IR session will finalize the entire process. Please let me know if you have any additional questions, or would like assistance confirming where the IR's disks are pointing."

I have confirmed our VM is running fully on the LUN we migrated it to.

[Gostev]

If you simply unpublish the VM, it will be deleted.

Unpublishing never deletes the VM... all it does is literally stops publishing virtual disks from backup file on vPower NFS share, that's it. As long as the VM is not dependent on those published virtual disks any longer (for example, their contents was migrated into new virtual disk files residing on production datastore, and VM runs off those), unpublishing will have no impact on the VM whatsoever. And, it does not matter if you used Quick Migration, or manual Storage VMotion.

[foggy]

Right, I might have been confused by the fact that in Hyper-V, VM is deleted on stopping IR session after being migrated via Live Migration.

[ibelyaev]

I'm new to Veeam, so the errors are forgivable
We use Vmware Vcenter 6.0, Vmware 6.0, 5.5 and Veeam Backup 9.0, individual VMware hosts without common storage, installed years ago.

We have a problem - 1 VM was corrupted, we wanted to restore VM from Veeam Backup, but it was 4.2 tb (2tb, 2tb, 200gb) and would be restored very long. I did an instant recovery on another host and success - the VM went up.
Then I read that to go back VM in production we have move it from NFS datastore created by Veeam to datastore of the host where it running. I did it not from Veeam Backup menu but via Vcenter migration.

Problem that VM is still running in Intant recovery mode in Veeam

Please help that to do
1. Migrate back VM to Veeam NFS volume and start recovery to production from Veeam menu
2. Other way cause VM already in other, not Veeam host

[veremin]

Since you've already migrated VM, using vCenter, there is no need to get it back to the "fake" Veeam datastore. Just click unpublish VM button in Instant VM Recovery wizard, the restore session will be closed, while migrated VM will stay intact. Thanks.

[ibelyaev]

Thank you. I checked all VM disks on datastore on new host. So I have to just unpublish IR of VM in Veeam.
Question: I made migration last weekend, will changes which was done on VM disks during the week be actual on VM after unpublishing.

[veremin]

Correct. Also, remember that you can always back up this VM with VeeamZIP prior to unpublishing it, just to be sure.