Strange effect of zfs rename

Willem Jan Withagen wjw at digiware.nl
Tue Aug 19 20:17:35 UTC 2014


On 18-8-2014 17:17, Ronald Klop wrote:
> On Sat, 09 Aug 2014 14:41:42 +0200, Willem Jan Withagen
> <wjw at digiware.nl> wrote:
> 
>> Hoi,
>>
>> In the process of creating backups I have this sequence of:
>>
>> /sbin/zfs destroy   zfsraid/backups at Saturday-2
>> /sbin/zfs rename -f zfsraid/backups at Saturday zfsraid/backups at Saturday-2
>> /sbin/zfs snapshot  zfsraid/backups at Saturday
>>
>> But then I end up with:
>> # cd /backups/.zfs/snapshot
>> # ll
>> ls: Saturday-2: Device busy
>> total 116
>>  0 dr-xr-xr-x  11 root  wheel  11 Aug  9 14:15 ./
>>  0 dr-xr-xr-x   4 root  wheel   4 Jul 30 22:07 ../
>> 17 drwxr-xr-x  13 root  wheel  16 Jul 31 09:54 20140801/
>> 17 drwxr-xr-x  18 root  wheel  21 Aug  5 01:03 Friday/
>> 17 drwxr-xr-x  16 root  wheel  19 Aug  3 01:03 Monday/
>> 17 drwxr-xr-x  15 root  wheel  18 Aug  2 01:03 Sunday/
>> 17 drwxr-xr-x  18 root  wheel  21 Aug  5 01:03 Thursday/
>> 17 drwxr-xr-x  17 root  wheel  20 Aug  4 01:03 Tuesday/
>> 17 drwxr-xr-x  18 root  wheel  21 Aug  5 01:03 Wednesday/
>> Exit 1
>>
>> And the device does not becomen not-busy even after a "long" wait.
>>
>> The way to remedy this is:
>> /sbin/zfs unmount -f zfsraid/backups
>> /sbin/zfs mount      zfsraid/backups
>>
>> After that the snapshot-dir is as it should.
>>
>> I'm running:
>> FreeBSD zfs.digiware.nl 9.3-STABLE FreeBSD 9.3-STABLE #272 r269145M: Sun
>> Jul 27 06:50:06 CEST 2014
>> root at zfs.digiware.nl:/usr/obj/usr/srcs/src9/src/sys/ZFS  amd64
>>
>> Can other reproduce this? Or is my system suffering from bitrot?
>>
>> Regards,
>> --WjW
> 
> I don't see this. I used to work on 9-STABLE and now on 10-STABLE (from
> august 5th).
> 
> I use zfs rename a lot for handling my snapshots, but I do not use the
> -f option. I must say I never need to go into the .zfs/snapshot
> directories, so they are mounted seldom. I just went into them today for
> a test, but I don't have 'Device busy' problems. Do you keep such a
> directory active somewhere?

No not really, it is like you suggest: rarely go into the snapshots for
recovery. But I do keep 'm online.

I'm just using test(1) to see if the snapshot I'd like to destroy is
really there. Which is easily done by testing if the snapshot dir exists.

--WjW




More information about the freebsd-fs mailing list