Superseeded-by:
https://lore.proxmox.com/pve-devel/20250729094114.66153-1-s.st...@proxmox.com/T/#u
On Mon Jul 28, 2025 at 4:43 PM CEST, Shannon Sterz wrote:
> zfs itself does not track the refquota per snapshot so we need handle
> this ourselves. this implementation tries to do so by leveraging a
> user property per snapshot.
>
> Signed-off-by: Shannon Sterz <s.st...@proxmox.com>
> ---
>
> this approach is not backward compatible, meaning that changes to volume
> sizes between snapshot before this patch will still be affected by this
> issue. however, it is fairly self-contained, does not require us to rely
> on the container config and works well with replication.
>
> we could fall back to resetting the refquota higher up in the call chain
> in case the storage doesn't manage to do it by itself. however, that
> comes with the potential downside of users messing with their configs
> und us resizing the disk when we shouldn't.
>
> src/PVE/Storage/ZFSPoolPlugin.pm | 46 +++++++++++++++++++++++++++++++-
> 1 file changed, 45 insertions(+), 1 deletion(-)
>
> diff --git a/src/PVE/Storage/ZFSPoolPlugin.pm
> b/src/PVE/Storage/ZFSPoolPlugin.pm
> index cdf5868..2474b7f 100644
> --- a/src/PVE/Storage/ZFSPoolPlugin.pm
> +++ b/src/PVE/Storage/ZFSPoolPlugin.pm
> @@ -482,9 +482,28 @@ sub volume_size_info {
> sub volume_snapshot {
> my ($class, $scfg, $storeid, $volname, $snap) = @_;
>
> - my $vname = ($class->parse_volname($volname))[1];
> + my (undef, $vname, undef, undef, undef, undef, $format) =
> $class->parse_volname($volname);
>
> $class->zfs_request($scfg, undef, 'snapshot',
> "$scfg->{pool}/$vname\@$snap");
> +
> + # if this is a subvol, track refquota information with snapshot, as zfs
> does
> + # not track this property via snapshosts and consequently does not roll
> it
> + # back
> + if ($format eq 'subvol') {
> + my $refquota = $class->zfs_request(
> + $scfg, undef, 'get', 'refquota', '-o', 'value', '-Hp',
> "$scfg->{pool}/$vname",
> + );
> +
> + chomp($refquota);
> +
> + $class->zfs_request(
> + $scfg,
> + undef,
> + 'set',
> + "pve-storage:refquota=${refquota}",
> + "$scfg->{pool}/$vname\@$snap",
> + );
> + }
> }
>
> sub volume_snapshot_delete {
> @@ -503,6 +522,31 @@ sub volume_snapshot_rollback {
>
> my $msg = $class->zfs_request($scfg, undef, 'rollback',
> "$scfg->{pool}/$vname\@$snap");
>
> + if ($format eq 'subvol') {
> + }
> +
> + # if this is a subvol, check if we tracked the refquota manually via
> user properties and if so,
> + # set it appropriatelly again
> + if ($format eq 'subvol') {
> + my $refquota = $class->zfs_request(
> + $scfg,
> + undef,
> + 'get',
> + 'pve-storage:refquota',
> + '-o',
> + 'value',
> + '-Hp',
> + "$scfg->{pool}/$vname\@$snap",
> + );
> +
> + chomp($refquota);
> +
> + if ($refquota =~ m/^\d+$/) {
> + $class->zfs_request($scfg, undef, 'set', "refquota=${refquota}",
> + "$scfg->{pool}/$vname");
> + }
> + }
> +
> # we have to unmount rollbacked subvols, to invalidate wrong kernel
> # caches, they get mounted in activate volume again
> # see zfs bug #10931 https://github.com/openzfs/zfs/issues/10931
> --
> 2.47.2
_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
