zfs itself does not track the refquota per snapshot so we need handle
this ourselves. otherwise rolling back a volume that has been resize
since the snapshot, will retain the new size. this is problematic, as
it means the value in the guest config does not longer match the size
of the disk on the storage.
this implementation tries to do so by leveraging a user property per
snapshot.
Reported-by: Lukas Wagner <l.wag...@proxmox.com>
Suggested-by: Fabian Grünbichler <f.gruenbich...@proxmox.com>
Signed-off-by: Shannon Sterz <s.st...@proxmox.com>
Reviewed-by: Fiona Ebner <f.eb...@proxmox.com>
---
Notes:
keeping fiona's R-b from v2 as v3 basically just incorporates her
suggestions (or suggestions she signed off on), but i'd drop the T-b
as these patches now behave somewhat differently (warning on unknown
values).
hope that reasoning is sound!
changes since v2, thanks @ Fiona Ebner:
- use variable for the snapshot name instead of constructing it
multiple times
- use `zfs_get_properties` to extract properties instead of
`zfs_request`
- warn on unknown refquota user property values
changes since v1:
- remove useless if statement
- reword commit message and comments to make them clearer
- add missing Suggested-by and Reported-by trailers
src/PVE/Storage/ZFSPoolPlugin.pm | 38 +++++++++++++++++++++++++++++---
1 file changed, 35 insertions(+), 3 deletions(-)
diff --git a/src/PVE/Storage/ZFSPoolPlugin.pm b/src/PVE/Storage/ZFSPoolPlugin.pm
index cdf5868..d8d8d0f 100644
--- a/src/PVE/Storage/ZFSPoolPlugin.pm
+++ b/src/PVE/Storage/ZFSPoolPlugin.pm
@@ -482,9 +482,25 @@ 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);
+ my $snapshot_name = "$scfg->{pool}/$vname\@$snap";
- $class->zfs_request($scfg, undef, 'snapshot',
"$scfg->{pool}/$vname\@$snap");
+ $class->zfs_request($scfg, undef, 'snapshot', $snapshot_name);
+
+ # if this is a subvol, track refquota information via user properties. zfs
+ # does not track this property for snapshosts and consequently does not
roll
+ # it back. so track this information manually.
+ if ($format eq 'subvol') {
+ my $refquota = $class->zfs_get_properties($scfg, 'refquota',
"$scfg->{pool}/$vname");
+
+ $class->zfs_request(
+ $scfg,
+ undef,
+ 'set',
+ "pve-storage:refquota=${refquota}",
+ $snapshot_name,
+ );
+ }
}
sub volume_snapshot_delete {
@@ -500,8 +516,24 @@ sub volume_snapshot_rollback {
my ($class, $scfg, $storeid, $volname, $snap) = @_;
my (undef, $vname, undef, undef, undef, undef, $format) =
$class->parse_volname($volname);
+ my $snapshot_name = "$scfg->{pool}/$vname\@$snap";
- my $msg = $class->zfs_request($scfg, undef, 'rollback',
"$scfg->{pool}/$vname\@$snap");
+ my $msg = $class->zfs_request($scfg, undef, 'rollback', $snapshot_name);
+
+ # if this is a subvol, check if we tracked the refquota manually via user
+ # properties and if so, set it appropriatelly again.
+ if ($format eq 'subvol') {
+ my $refquota = $class->zfs_get_properties($scfg,
'pve-storage:refquota', $snapshot_name);
+
+ if ($refquota =~ m/^\d+$/) {
+ $class->zfs_request(
+ $scfg, undef, 'set', "refquota=${refquota}",
"$scfg->{pool}/$vname",
+ );
+ } elsif ($refquota ne "-") {
+ # refquota user property was set, but not a number -> warn
+ warn "property for refquota tracking contained unknown value
'$refquota'\n";
+ }
+ }
# we have to unmount rollbacked subvols, to invalidate wrong kernel
# caches, they get mounted in activate volume again
--
2.47.2
_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
