[pve-devel] [PATCH storage] zfs plugin: alloc/clone: cleanup image when creating LU mapping fails

[Christoph Heiss]

Tested this by setting up a iSCSI target using targetcli(d) on a
separate PVE 8.4 system as a base (due to ZFS goodies) and then adding a
ZFS-over-iSCSI storage using the LIO provider to a test cluster.

Due to #5071 [0], this configuration is currently broken and allocation
will fail.

I then tried to add a new hard disk to a VM on the ZFS-over-iSCSI
storage, which
a) failed as expected and logs the error
b) the created LUN was cleaned up correctly. Also verified on the target
   that the previously created zvol is indeed cleaned up.

Actual changes are simple enough & look good, so please consider this
patch:

Tested-by: Christoph Heiss <c.heiss at proxmox.com>
Reviewed-by: Christoph Heiss <c.heiss at proxmox.com>

[0] https://bugzilla.proxmox.com/show_bug.cgi?id=5071

On Thu Jun 5, 2025 at 1:57 PM CEST, Fiona Ebner wrote:
> Avoid leaving behind orphaned images if creating the associated LU
> mapping fails during image allocation or cloning.
>
> Reported-by: Christoph Heiss <c.heiss at proxmox.com>
> Signed-off-by: Fiona Ebner <f.ebner at proxmox.com>
> ---
>  src/PVE/Storage/ZFSPlugin.pm | 27 ++++++++++++++++++++++-----
>  1 file changed, 22 insertions(+), 5 deletions(-)
>
> diff --git a/src/PVE/Storage/ZFSPlugin.pm b/src/PVE/Storage/ZFSPlugin.pm
> index 94cb11f..4118aa4 100644
> --- a/src/PVE/Storage/ZFSPlugin.pm
> +++ b/src/PVE/Storage/ZFSPlugin.pm
> @@ -6,6 +6,7 @@ use IO::File;
>  use POSIX;
>  use PVE::Tools qw(run_command);
>  use PVE::Storage::ZFSPoolPlugin;
> +use PVE::RESTEnvironment qw(log_warn);
>  use PVE::RPCEnvironment;
>
>  use base qw(PVE::Storage::ZFSPoolPlugin);
> @@ -283,8 +284,16 @@ sub clone_image {
>      # get ZFS dataset name from PVE volname
>      my (undef, $clonedname) = $class->parse_volname($name);
>
> -    my $guid = $class->zfs_create_lu($scfg, $clonedname);
> -    $class->zfs_add_lun_mapping_entry($scfg, $clonedname, $guid);
> +    eval {
> +	my $guid = $class->zfs_create_lu($scfg, $clonedname);
> +	$class->zfs_add_lun_mapping_entry($scfg, $clonedname, $guid);
> +    };
> +    if (my $err = $@) {
> +	print "cleaning up allocated image '$clonedname' after failure to create LU mapping\n";
> +	eval { $class->zfs_delete_zvol($scfg, $clonedname); };
> +	log_warn("cleanup failed - $@") if $@;
> +	die "error while creating LU mapping was - $err";
> +    }
>
>      return $name;
>  }
> @@ -302,9 +311,17 @@ sub alloc_image {
>      $volname = $class->find_free_diskname($storeid, $scfg, $vmid, $fmt) if !$volname;
>
>      $class->zfs_create_zvol($scfg, $volname, $size);
> -
> -    my $guid = $class->zfs_create_lu($scfg, $volname);
> -    $class->zfs_add_lun_mapping_entry($scfg, $volname, $guid);
> +
> +    eval {
> +	my $guid = $class->zfs_create_lu($scfg, $volname);
> +	$class->zfs_add_lun_mapping_entry($scfg, $volname, $guid);
> +    };
> +    if (my $err = $@) {
> +	print "cleaning up allocated image '$volname' after failure to create LU mapping\n";
> +	eval { $class->zfs_delete_zvol($scfg, $volname); };
> +	log_warn("cleanup failed - $@") if $@;
> +	die "error while creating LU mapping was - $err";
> +    }
>
>      return $volname;
>  }