[pve-devel] [storage] Lack of deactivate_volume call after cloning from a snapshot

[Fabian Grünbichler]

On October 1, 2025 12:39 am, Andrei Perapiolkin via pve-devel wrote:
> Hi,
> 
> 'deactivate_volume' in the storage plugin is not called after cloning
> completes if the clone is initiated from the Proxmox web UI and both VMs
> are on the same host.
> 
> I’ve created a bug report: https://bugzilla.proxmox.com/show_bug.cgi?id=6879
> 
> In my understanding, the problem is related to:
> 
> /method({ name => 'clone_vm',  path => '{vmid}/clone', ...
> /
> Located in PVE/API2/Qemu.pm around 1459 - 4568
> 
> In this API method, deactivation calls are triggered only when the
> $target parameter is provided.
> 
>                  if ($target) {
>                      if (!$running) {
>                          # always deactivate volumes - avoids that LVM
> LVs are active on several nodes
>                          eval {
> PVE::Storage::deactivate_volumes($storecfg, $vollist, $snapname);
>                          };
>                          # but only warn when that fails (e.g., parallel
> clones keeping them active)
>                          log_warn($@) if $@;
>                      }
> 
>                      PVE::Storage::deactivate_volumes($storecfg,
> $newvollist);
> 
>                      my $newconffile =
> PVE::QemuConfig->config_file($newid, $target);
>                      die "Failed to move config to node '$target' -
> rename failed: $!\n"
>                          if !rename($conffile, $newconffile);
>                  }
> 
> https://github.com/proxmox/qemu-server/blob/31d6f5f63bd6edb6c43de3e0213d38199c350cd4/src/PVE/API2/Qemu.pm#L4533C1-L4547C18
> 
> It seems that web UI, for cloning on the same node, does not provide
> '$target' and as a result no deactivation being triggered volume
> snapshot and target volume.
> 
> Is it expected behavior?
> 
> Will it be OK to move 'deactivate' section outside of the $target check?

also filed as https://bugzilla.proxmox.com/show_bug.cgi?id=6879, see my
reply there:

the issue is that multiple such clones can run in parallel, and there is
no ref counting at the moment that would make deactivation safe in this
scenario. that is also the reason why we don't deactivate volumes in
other places where it's impossible to tell whether there are no more
other users.