[Gluster-users] Quota limits gone after upgrading to 3.8

[Sanoj Unnikrishnan]

Hi Mabi,

Note that limits are still configured and working.
re-adding the limits will not help here (unless you are willing to disable
and re-enable quota first).
The reason is if a gfid exists in quota.conf (because a limit was earlier
set on it), it does not  need change when limit changes.
The quota.conf file only keep track of which gfid have limit set. The
original value of the limits are set in xattr on filesystem

Another work around without mauallly touching quota.conf is,
> Create a new dummy directory anywhere in the FS. add a limit in this
directory.

After this you should be able to see the listing.
If you remove this dummy directory or limit on it, you will once again be
exposed to same issue.

Regards,
Sanoj

On Tue, May 9, 2017 at 10:59 PM, mabi <mabi at protonmail.ch> wrote:

> Hi Sanoj,
>
> Thanks for pointing me at this bug, I was not aware about it.
>
> As this is a production GlusterFS cluster I would rather not mess with the
> quota.conf file as you suggested. Instead I will simply re-add all my
> quotas by running the following command again:
>
> gluster volume quota myvolume limit-usage /directory1 100GB
>
> Can you confirm me that this is safe to run again?
>
>
>
> As soon as I have a minute I will complete your survey about quotas.
>
> Best,
> M.
>
> -------- Original Message --------
> Subject: Re: [Gluster-users] Quota limits gone after upgrading to 3.8
> Local Time: May 9, 2017 6:50 AM
> UTC Time: May 9, 2017 4:50 AM
> From: sunnikri at redhat.com
> To: mabi <mabi at protonmail.ch>
> Gluster Users <gluster-users at gluster.org>
>
> Hi mabi,
>
> This bug was fixed recently, https://bugzilla.redhat.com/sh
> ow_bug.cgi?id=1414346. It would be available in 3.11 release. I will plan
> to back port same to earlier releases.
>
> Your quota limits are still set and honored, It is only the listing that
> has gone wrong. Using list with command with single path should display the
> limit on that path. The printing of list gets messed up when the last gfid
> in the quota.conf file is not present in the FS (due to an rmdir without a
> remove limit)
>
> You could use the following workaround to get rid of the issue.
>  => Remove exactly the last 17 bytes of " /var/lib/glusterd/vols/<volnam
> e>/quota.conf"
>       Note: keep a backup of quota.conf for safety
> If this does not solve the issue, please revert back with
> 1) quota.conf file
> 2) output of list command (when executed along with path)
> 3) getfattr -d -m . -e hex <path_on_brick _where_limit_is_set> | grep limit
> It would be great to have your feedback for quota on this thread (
> http://lists.gluster.org/pipermail/gluster-users/2017-April/030676.html)
>
> Thanks & Regards,
> Sanoj
>
>
> On Mon, May 8, 2017 at 7:58 PM, mabi <mabi at protonmail.ch> wrote:
>
>> Hello,
>>
>> I upgraded last week my 2 nodes replica GlusterFS cluster from 3.7.20 to
>> 3.8.11 and on one of the volumes I use the quota feature of GlusterFS.
>> Unfortunately, I just noticed by using the usual command "gluster volume
>> quota myvolume list" that all my quotas on that volume are gone. I had
>> around 10 different quotas set on different directories.
>>
>> Does anyone have an idea where the quotas have vanished? are they gone
>> for always and do I need to re-set them all?
>>
>> Regards,
>> M.
>>
>>
>>
>> _______________________________________________
>> Gluster-users mailing list
>> Gluster-users at gluster.org
>> http://lists.gluster.org/mailman/listinfo/gluster-users
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20170510/9cde4f72/attachment.html>