[automated-testing] Bringing up bricks with/without shd

[Vijay Bhaskar Reddy Avuthu]

The main intention of test case is heal shouldn't happen when shd is not
running on source brick. I think we can disable self-heal daemon to achieve
the similar behaviour

On Thu, May 31, 2018 at 10:39 AM, Nigel Babu <nigelb at redhat.com> wrote:

>
>
> On Mon, May 28, 2018 at 9:15 AM, Atin Mukherjee <amukherj at redhat.com>
> wrote:
>
>>
>>
>> On Mon, May 28, 2018 at 8:50 AM, Nigel Babu <nigelb at redhat.com> wrote:
>>
>>> (Adding Atin)
>>>
>>> We're making an assumption that when bricks are brought online, SHD
>>> isn't brought online. Did this behavior change recently?
>>>
>>
>> AFAIK, there's no change of behaviour as such which got introduced
>> recently.
>>
>>
>>> Atin, is there a difference to how bricks should be killed with brick
>>> multiplex on vs off?
>>>
>>
>> For non brick multiplexing cases, sending a SIGTERM (kill -15)  to a
>> process is the way to go, however with brick multiplexing following the
>> same approach would eventually kill the entire process resulting in all
>> other brick instances (attached bricks) to be also going down which is
>> unexpected. Like our upstream regression test infrastructure we should use
>> gf_attach () utility to kill any bricks w/ brick multiplexing.
>>
>>
>>> On Fri, May 25, 2018 at 4:38 PM, Vitalii Koriakov <vkoriako at redhat.com>
>>> wrote:
>>>
>>>> Hello all.
>>>>
>>>> We have such scenario:
>>>> - brink brick offline
>>>> - kill shd processes on other bricks
>>>> - bring killed brick online
>>>>
>>>> The thing is that with bringing brick, we bring up shd processes on
>>>> other bricks also.
>>>> Its connected with brick mux, so, killing brick and bringing brick is
>>>> different and it seems that brickmux is enabled.
>>>> So we have failed test, because of such behavior.
>>>>
>>>> We need a mechanism to bring brick up without shd.
>>>>
>>>
>> Not sure how this test is a requirement. The only way to bring up brick
>> processes is to do a volume start force and with that all the respective
>> daemons (including shd) also comes up.
>>
>>
>>>>
> Vijay, why is this a test requirement? From Atin's response, it sounds
> like Gluster is behaving as expected here.
>
>
> --
> nigelb
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/automated-testing/attachments/20180531/18894213/attachment-0002.html>