Open Stack

On Wed, Feb 14, 2018 at 6:45 AM, Andreas Jaeger <aj at suse.com> wrote:
> On 2018-02-14 01:28, Ghanshyam Mann wrote:
>>  On Wed, Feb 14, 2018 at 12:06 AM, Paul Belanger 
>> <pabelanger at redhat.com> wrote:
>>>  On Tue, Feb 13, 2018 at 11:05:34PM +0900, gmann wrote:
>>>>  Hi Infra Team,
>>>> 
>>>>  I have 1 quick question on zuulv3 jobs and their migration part. 
>>>> From
>>>>  zuulv3 doc [1], it is clear about migrating the job definition 
>>>> and use
>>>>  those among cross repo pipeline etc.
>>>> 
>>>>  But I did not find clear recommendation that whether project's
>>>>  pipeline definition should stay in project-config or we should 
>>>> move
>>>>  that to project side.
>>>> 
>>>>  IMO,
>>>>  'template' part(which has system level jobs) can stay in
>>>>  project-config. For example below part-
>>>> 
>>>>  
>>>> https://github.com/openstack-infra/project-config/blob/e2b82623a4ab60261b37a91e311118301927b9b6/zuul.d/projects.yaml#L10507-L10523
>>>> 
>>>>  Other pipeline definition- 'check', 'gate', 'experimental' etc 
>>>> should
>>>>  be move to project repo, mainly this list-
>>>>  
>>>> https://github.com/openstack-infra/project-config/blob/master/zuul.d/projects.yaml#L10524-L11019
>>>> 
>>>>  If we move those past as mentioned above then, we can have a
>>>>  consolidated place to control the project pipeline for
>>>>  'irrelevant-files', specific branch etc
>>>> 
>>>>  ..1 https://docs.openstack.org/infra/manual/zuulv3.html
>>>> 
>>>  As it works today, pipeline stanza needs to be in a config 
>>> project[1]
>>>  (project-config) repo. So what you are suggestion will not work. 
>>> This was done
>>>  to allow zuul admins to control which pipelines are setup / 
>>> configured.
>>> 
>>>  I am mostly curious why a project would need to modify a pipeline 
>>> configuration
>>>  or duplicate it into all projects, over having it central located 
>>> in
>>>  project-config.
>> 
>>  pipeline stanza  and configuration stay in project-config. I mean 
>> list
>>  of jobs defined in each pipeline for specific project for example
>>  here[2]. Now we have list of jobs for each pipeline in 2 places, one
>>  in project-config [2] and second in project repo[3].
>> 
>>  Issue in having it in 2 places:
>>  - No single place to check what all jobs project will run with what 
>> conditions
>>  - If we need to modify the list of jobs in pipeline or change other
>>  bits like irrelevant-files etc then it has to be done in
>>  project-config. So no full control by project side.
> 

For me it is even more than two places as the project templates like 
'integarted-gate'[4] defines jobs to be executed on a project that 
includes the template in the project-config. Which leads to problems 
like [5]. This shows that tracking down why some job runs on a change 
is fairly non-trivial from a developer perspective. Therefore I support 
to define which jobs run on a given project as close to the project as 
possible and as small number of different places as possible. I even 
volunteer to help with the moving from nova perspective.


> This should be explained in:
> https://docs.openstack.org/infra/manual/zuulv3.html#what-to-convert
> 
> So, the standard templates/jobs - incl. PTI mandated ones - should 
> stay
> in project-config, you can move everything else in-tree,

As far as I understand this list allows us to solve [5] by simply 
moving every jobs from 'integrated-gate' to the respective project in 
tree as the jobs in that template are not part of the PTI.


[4] 
https://github.com/openstack-infra/openstack-zuul-jobs/blob/df8a8e8ee41c1ceb4da458a8681e39de39eafded/zuul.d/zuul-legacy-project-templates.yaml#L93
[5] https://review.openstack.org/#/c/538908

Cheers,
gibi