/.zfs/ directory works differently on Linux - possible fix?

[Former user]

On linux, it seems that rather than a single .zfs directory, there's one per fs which list each filesystem's snapshots, so getting rid of expired snapshots doesn't work.

Rather than ls /.zfs/snapshot/, you could try the following:

# prune dbs if requested
if [ "x$maxnum" != x ]
then
    zfs list -H -t snapshot -s creation -o name | grep ^$zpool\@$label_pfx- | awk "NR > $maxnum {print}" | while read snapname
    do
        zfs destroy -r $snapname || exit $?
    done
fi

This works nicely on Ubuntu and I think it should work on BSD too, but obviously would benefit from a little testing!

[cdnt]

Spotted a typo so signed up to comment - it should be "-S creation" with an uppercase S, not lowercase...

[francoisstark]

On Ubuntu 14.04 the /.zfs directory did not exits, but I found the snapshots listed under /[poolname]/.zfs/snapshot/, so I just change line 77 to be:

ls /$zpool/.zfs/snapshot/ ....

which seems to be a smaller change from the original code.

[cdnt]

...but using ls will not work if the mount point is changed (mine is). The version using zfs list works wherever a pool or fs is mounted. And should also work across operating systems or if someone has set listsnapshots=off. It might be a bigger change, but is more robust.

[Willard Korfhage]

Alternatively, you can replace

ls /.zfs/snapshot

with

ls zfs get -H -o value mountpoint $zpool/.zfs/snapshot/