Expired Snapshots Not Being Removed

[Fiddlehead]

I'm trying to get a replication task set up in v9.3. I don't plan on storing snapshots though. The NAS is only holding daily backups of another filesystem, which has so much turnover as to make snapshots very bulky and impractical. I only need a stable image of the filesystem to replicate to a remote secondary device; it can then be promptly removed from primary FreeNAS.

Right now, I have daily snapshot task with a two hour window, with snapshots set to expire after 10 hours, followed by a replication task two hours later. As I understand it, the schedule would be:
22:00 Take a snapshot - auto-20150101.2000-12h
24:00 Replicate auto-20150101.2000-12h to secondary
8:00 Snapshot auto-20150101.2000-12h expires
22:00 Take new snapshot auto-20150102.2000-12h
22:xx Remove snapshot auto-20150101.2000-12h
24:00 Replicate auto-20150102.2000-12h to secondary

Snapshots are taken and replicated as expected, but old snapshots still remain behind. So after a few full filesystem turnovers, the primary NAS gets completely filled.

During the snapshot operation, I'm seeing a new snapshot created, and a hold put on it; followed by this line in var/log/messages every minute for the rest of the snapshot window:

Code:

srvrbkp01 autosnap.py: [tools.autosnap:58] Popen()ing: /sbin/zfs get -
H freenas:state zpool1/nas_bkp_1@auto-20150101.2000-12h

I haven't come across any errors, just that operation repeated on the expired snapshot until the snapshot window ends and the replication starts.

Is there something I'm not understanding here about how snapshots are cleaned up?

 

[dlavigne]

Which build version (from System -> Information)?

 

[Fiddlehead]

FreeNAS-9.3-STABLE-201502271818

 

[dlavigne]

There was a bug that prevented expired snapshots from being deleted. I'm pretty sure that has been fixed. Does upgrading to the latest SU fix it?

 

[Fiddlehead]

Hmm... apparently the bug you refer to didn't appear until 9.3.1, and the fix is staged for the next release. I'll wait until it's released and see if it fixes my problem anyway. In the meantime, I've created a cron-based workaround to remove replication holds and delete snapshots before the snapshot and replication tasks are run:

Code:

0 18 * * * zfs list -H -t snapshot -o name | grep nas_bkp | xargs -L 1 zfs release -r freenas
:repl ; zfs list -H -t snapshot -o name | grep nas_bkp | xargs -L 1 zfs destroy -rd

Do cron jobs stick around in FreeNAS? I know the GUI likes overwriting things...

 

[Bidule0hm]

You can setup your cron jobs in the GUI ;)

 

[Fiddlehead]

Thanks! I didn't notice that. Added via GUI instead.