Hello,

Not sure IO could be the case. The underlying storage itself is brand new (nvme) connected with FC and is barely at 10 % capacity with low IOPS and practically zero latency. There are no IO limitations on the LUN itself. I would also be able to see any IO problems on the other VMs (none in this case).

I'm out of ideas on what to do for the time how to stop / complete the task. Any suggestion welcome.

Shutting down the VM in this state means that It probably wont start back (snapshots with disks in illegal state). Worst case I plan to restore this VM from a backup tonight.

Regards,

On Wed, May 27, 2020 at 4:39 PM Benny Zlotnik <bzlotnik@redhat.com> wrote:

Sorry, by overloaded I meant in terms of I/O, because this is an
active layer merge, the active layer
(aabf3788-8e47-4f8b-84ad-a7eb311659fa) is merged into the base image
(a78c7505-a949-43f3-b3d0-9d17bdb41af5), before the VM switches to use
it as the active layer. So if there is constantly additional data
written to the current active layer, vdsm may have trouble finishing
the synchronization

On Wed, May 27, 2020 at 4:55 PM David Sekne <david.sekne@gmail.com> wrote:
>
> Hello,
>
> Yes, no problem. XML is attached (I ommited the hostname and IP).
>
> Server is quite big (8 CPU / 32 Gb RAM / 1 Tb disk) yet not overloaded. We have multiple servers with the same specs with no issues.
>
> Regards,
>
> On Wed, May 27, 2020 at 2:28 PM Benny Zlotnik <bzlotnik@redhat.com> wrote:
>>
>> Can you share the VM's xml?
>> Can be obtained with `virsh -r dumpxml <vm_name>`
>> Is the VM overloaded? I suspect it has trouble converging
>>
>> taskcleaner only cleans up the database, I don't think it will help here
>>