[gluster-packaging] [Gluster-Maintainers] 3.7.10 known issue

Kaushal M kshlmster at gmail.com
Tue Apr 5 05:49:11 UTC 2016


On Tue, Apr 5, 2016 at 9:23 AM, Atin Mukherjee <amukherj at redhat.com> wrote:
>
>
> On 04/04/2016 11:10 PM, Niels de Vos wrote:
>> On Mon, Apr 04, 2016 at 08:19:30PM +0530, Atin Mukherjee wrote:
>>>
>>>
>>> On 04/04/2016 07:43 PM, Atin Mukherjee wrote:
>>>>
>>>>
>>>> On 04/04/2016 06:53 PM, Kaleb KEITHLEY wrote:
>>>>> On 04/04/2016 09:11 AM, Kaleb KEITHLEY wrote:
>>>>>> On 04/04/2016 07:41 AM, Niels de Vos wrote:
>>>>>>> On Fri, Apr 01, 2016 at 09:56:45AM +0530, Atin Mukherjee wrote:
>>>>>>>> Hi Kaushal,
>>>>>>>>
>>>>>>>> We hit a regression yesterday where GlusterD fails to restart when the
>>>>>>>> volume has snapshots. Refer to [1] for more details. The work around is
>>>>>>>> also a bit tedious where each snapshot brick folders have to be created
>>>>>>>> manually and I don't think see any other alternative workaround here.
>>> One correction here, the issue is seen only when the node reboots, not
>>> glusterd. Considering that users won't reboot machines often, we can
>>> still live with it for the next scheduled release?
>>
>> No, I do not think that is acceptible. Many users will install updates
>> and do a reboot after that. Even having just a few users rebooting and
>> rendering their deployment unusable is a huge problem.
> Ok, but I was wondering how many users use snapshot? I've seen almost no
> emails related to snapshot issues in gluster-users mailing list. Anyways
> I don't have any issue in releasing 3.7.11 early.

I've got no problem with doing 3.7.11 early.
I've not announced 3.7.10 yet, so we could just announce 3.7.11 instead.

This should serve as a catalyst to get started on our planned
automated release tests.
I did to some manual testing before tagging, but I didn't cover
restarts, or snapshots.
Covering everything, every time in manual testing is not feasible.


>>
>> Niels
>>
>>> I'll be working on
>>> the release note to put down the work around.
>>>>>>>>
>>>>>>>> IMO, we should mention this as known issue in the release note for 3.7.10.
>>>>>>>
>>>>>>> This sounds serious enough to not package 3.7.10 and have users update
>>>>>>> to it automatically.
>>>>>>
>>>>>> That djinni is already out of the bottle.
>>>>>
>>>>> If there's a fix available then it seems (to me) that a better path
>>>>> forward is to release 3.7.11 quickly.
>>>> Patch http://review.gluster.org/#/c/13869 is under review.
>>>>>
>>>>> --
>>>>>
>>>>> Kaleb
>>>>>
>>>>>
>>>> _______________________________________________
>>>> maintainers mailing list
>>>> maintainers at gluster.org
>>>> http://www.gluster.org/mailman/listinfo/maintainers
>>>>
>>> _______________________________________________
>>> maintainers mailing list
>>> maintainers at gluster.org
>>> http://www.gluster.org/mailman/listinfo/maintainers
> _______________________________________________
> maintainers mailing list
> maintainers at gluster.org
> http://www.gluster.org/mailman/listinfo/maintainers


More information about the packaging mailing list