Re: Again, no space left on device while rebalancing and recipe doesnt work

[Qu Wenruo <quwenruo@cn.fujitsu.com>]

Marc Haber wrote on 2016/02/29 16:33 +0100:
> Hi,
>
> On Mon, Feb 29, 2016 at 09:56:58AM +0800, Qu Wenruo wrote:
>> Marc Haber wrote on 2016/02/27 22:14 +0100:
>>> I have again the issue of no space left on device while rebalancing
>>> (with btrfs-tools 4.4.1 on kernel 4.4.2 on Debian unstable):
>>>
>>> mh@fan:~$ sudo btrfs balance start /mnt/fanbtr
>>> ERROR: error during balancing '/mnt/fanbtr': No space left on device
>>
>> It seems that, only when balancing all chunks, ENOSPC error happens.
>>
>> And did you run any other heavy IO at background?
>
> Not when running those last commands for the mailing list post.
>
>> BTW, is there any kernel log when the ENOSPC happens?
>
>> Would you please try the following commands to see which one caused the
>> problem?
>> And would you please provide the dmesg of them?
>>
>> # btrfs balance start -dprofiles=single /mnt/fanbtr
>> # btrfs balance start -mprofile=dup /mnt/fanbtr
>> # btrfs balance start -sprofile=dup /mnt/fanbtr
>
> I have attached the logs. I used logger(1) to have in syslog which
> command I executed, and I have piped the userspace's output to logger
> so that the syslog entries match the userspace output.

Didn't see the attachment though, seems to be filtered by maillist police.

>
> -mprofile gave an error message, I therefore tried -mprofiles, and
> -sprofiles wanted me to use the --force, so I did that as well.
>
> The three balance commands above all three finshed alright without
> running into ENOSPC, while running a plain balance (which is also part
> of the log) errors out every time.

Strange, but at least some small clue to chase.

>
> And, the -dprofiles=single log caused a number of INFOs regarding
> btrfs-cleaner and btrfa-balance processes gotten stuck for more than
> 120 seconds during the run.

That's not normal and it would be a bug.

>
> I now have a kworker and a btfs-transact kernel process taking most of
> one CPU core each, even after the userspace programs have terminated.
> Is there a way to find out what these threads are actually doing?

Did btrfs balance status gives any hint?

Thanks,
Qu

>
> Greetings
> Marc
>