Virtual Machine Unresponsive on Snapshot Removal (30 seconds+) when CBT enabled

Hi,

I have a couple virtual machines on a NFS datastore running on a NetApp Filer. Everything works well normally. The storage is not overloaded.

On my regular virtual machines, never backed up using VMware Data Recovery, a snapshot removal takes only a couple seconds and the machine stays unresponsive for a very small/insignificant time.

But after I do a backup with VMware Data Recovery, the snapshot removal process starts to take 30 seconds or more. And this is for all snapshot removals on the VM. The Virtual Machine becames unresponsive for that time, losing all kind of network connectivity and even the Vi-Client console stalls.

If I move the virtual machine to local storage, the snapshot removal is again fast. Moving it back to the NFS datastore makes it slow again.

I tracked the problem down to the Change Block Tracking (CTK) feature.

VMware Data Recovery adds a ctkEnabled = true, scsi0:X:ctkEnabled = true  to the Virtual Machine advanced configuration when it perform a backup of it.

As soon as I set these entries to false, the Virtual Machine snapshot removal works as it should. It just takes some seconds to consolidate the snapshot.

If I backup the virtual machine again with VDR, the erratic behavior shows up again, because VDR resets the values to "true" again.

I can reproduce the problem using the following steps:

1. Create or clone a virtual machine, using an NFS datastore as the vmdk target (in my case, on a NetApp FAS2020).

2. Boot the machine (don't need an OS for the problem to show up) 3. Take a snapshot. Wait and remove a snapshot. It should take some seconds 4. Backup the machine using VDR 5. Take a snapshot and remove it (same as step 3).  The problem shows up!!

Result of the procedure:

- Snapshot removal takes 30 seconds or more on step 5. Considerably more time than step 3 (before VDR backup).

- If you have an OS on the machine, ping the machine during snapshot removal on step 5. The machine is unresponsive during the snapshot removal operation.

Is anyone having the same problem?

Any idea what can be causing this?

Thanks.