zfs destroy snapshot doesn't free space

[Jeremy Chadwick]

On Fri, Aug 13, 2010 at 04:51:24PM +0200, Andreas Mayer wrote:
> I have a problem with my ZFS storage: some application filled a
> certain directory in /var completely up with data and the server runs
> a script which takes a snapshot every night. So, ~650 GB of the
> available 700 GB were filled up.
> 
> Then I destroyed the last two snapshots (each referencing about 300
> GB) and the files on the live system so that now there are no
> snapshots for
> /var and "du -hs /var" reports a size of 2 GB.
> 
> However, zfs still reports that 623G are referenced:
> 
> # zfs list rpool/var
> NAME        USED  AVAIL  REFER  MOUNTPOINT
> rpool/var   623G  48.4G   623G  /var
> 
> # zfs get all rpool/var
> ...
> rpool/var  usedbysnapshots       0                      -
> rpool/var  usedbydataset         623G                   -
> rpool/var  usedbychildren        0                      -
> rpool/var  usedbyrefreservation  0                      -
> ...
> 
> If I take a snapshot again, this snapshot also references 623G.
> 
> What can I do to reclaim this space? I have to do this before I can
> set a quota (I have set quotas for all other file systems now :) ).

Can you provide uname -a output please?  Thanks.

-- 
| Jeremy Chadwick                                   jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.              PGP: 4BD6C0CB |