Re: [openstack-dev] [all][infra] NOT_REGISTERED to be or not to be

[Paul Belanger]

On Mon, Aug 01, 2016 at 03:53:12PM +, Lenny Verkhovsky wrote:
> Hi,
> 
> Currently in some cases[1] CI sets a comment on the patch set as 
> NOT_REGISTERED.
> 
> Those comments are very hard to monitor for CI operators and have noisy value 
> for the developers.
> 
> Maybe a better solution is not commenting in such cases at all as discussed 
> in [2].
> 
> If a developer is missing some important CI comments, it can rechecked later 
> or sent an email to CI owner.
> 
> [1] No valid slaves
>Not all jobs are registered due to zuul restart for instance
> 
> [2] 
> http://eavesdrop.openstack.org/irclogs/%23openstack-meeting/%23openstack-meeting.2016-08-01.log.html
> 
We actually disabled this functionality in openstack-infra by adding a new
setting to zuul called check_job_registration[3]. While it doesn't address the
issue of invalid job configuration from landing in zuul, it does prevent
NOT_REGISTERED in the gate now.

We, openstack-infra, make every effort in our gate testing to ensure each change
to jenkins/jobs and zuul/layout.yaml is actually valid, if not zuul with -1 the
change as invalid. In our case, having zuul then check for registered jobs, was
a little redundant.

Now, changes land, zuul skips checking if the jobs are valid, and waits for new
nodes to come online. It also is less disruptive to testing when we have to
restart zuul, since jobs does fail with NOT_REGISTERED.

The downside of course, if a job is not configured properly, it will linger in
the queue until somebody address the issue.

[3] 
http://git.openstack.org/cgit/openstack-infra/zuul/commit/?id=9208dc1c0a859642deece4f4be5f43fae065c945

__
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

Re: [openstack-dev] [all][infra] NOT_REGISTERED to be or not to be

[Mikhail Medvedev]

Thanks starting the discussion, Lenny. I thought about a way to
accomplish that, and came up with few options:

 - Pre-register all possible jenkins jobs with gearman. This only
avoids not registered errors due to gearman server restart, not if you
misconfigured your system;

 - Add option to zuul to treat builds as effectively canceled if there
is no corresponding gearman worker, this would avoid the
not_registered comment in all cases. This can be done in one line of
code if you just hack it [1] (assuming my patch is correct).

 - Automatically register all jobs that zuul tries to start with
gearman. That is, check if  exists on gearman, and register
dummy function if not, for each attempted build. This would avoid
missing any patches.

[1] 
https://review.openstack.org/#q,Ie6d5ea35c6eeed465168f24921b04442df8f5744,n,z


On Mon, Aug 1, 2016 at 10:53 AM, Lenny Verkhovsky  wrote:
> Hi,
>
>
>
> Currently in some cases[1] CI sets a comment on the patch set as
> NOT_REGISTERED.
>
>
>
> Those comments are very hard to monitor for CI operators and have noisy
> value for the developers.
>
>
>
> Maybe a better solution is not commenting in such cases at all as discussed
> in [2].
>
>
>
> If a developer is missing some important CI comments, it can rechecked later
> or sent an email to CI owner.
>
>
>
> [1] No valid slaves
>
>Not all jobs are registered due to zuul restart for instance
>
>
>
> [2]
> http://eavesdrop.openstack.org/irclogs/%23openstack-meeting/%23openstack-meeting.2016-08-01.log.html
>
>
>
>
>
> Thanks.
>
> Lenny
>
>
> __
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>

__
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

[openstack-dev] [all][infra] NOT_REGISTERED to be or not to be

[Lenny Verkhovsky]

Hi,

Currently in some cases[1] CI sets a comment on the patch set as NOT_REGISTERED.

Those comments are very hard to monitor for CI operators and have noisy value 
for the developers.

Maybe a better solution is not commenting in such cases at all as discussed in 
[2].

If a developer is missing some important CI comments, it can rechecked later or 
sent an email to CI owner.

[1] No valid slaves
   Not all jobs are registered due to zuul restart for instance

[2] 
http://eavesdrop.openstack.org/irclogs/%23openstack-meeting/%23openstack-meeting.2016-08-01.log.html


Thanks.
Lenny
__
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev