Actively triggering of CI jobs

Wed Dec 9 09:35:14 UTC 2015

On 12/09 11:10, Amit Aviram wrote:
> On Wed, Dec 9, 2015 at 10:05 AM, Eyal Edri <eedri at redhat.com> wrote:
> 
> > +1,
> > this will be the best suggestion.
> > we can try adding a manual trigger for drafts if needed, still need to
> > check if possible.
> >
> > e.
> >
> > On Wed, Dec 9, 2015 at 9:38 AM, Eli Mesika <emesika at redhat.com> wrote:
> >
> >>
> >>
> >> ----- Original Message -----
> >> > From: "Barak Korren" <bkorren at redhat.com>
> >> > To: "Amit Aviram" <aaviram at redhat.com>
> >> > Cc: "infra" <infra at ovirt.org>
> >> > Sent: Tuesday, December 8, 2015 5:16:30 PM
> >> > Subject: Re: Actively triggering of CI jobs
> >> >
> >> > > I was thinking, maybe it would be better if we will explicitly
> >> require to
> >> > > run the CI jobs when we push patches.. then only when the developer
> >> will
> >> > > need the job's feedback it will be activated. no redundant jobs will
> >> run,
> >> > > and we will wait much less for the jobs to finish when we will
> >> actually
> >> > > need
> >> > > them.
> >>
> >> Why not simply submit your patches as a Draft until the point you want CI
> >> to run on them, then you can simply publish them ...
> >> This is the way I am using and it's simple ...
> >>
> >> > >
> >> > It seems to me that it will me too easy to forget to run the CI this
> >> way.
> >>
> > We barely merge patches that did not pass the CI tests.. only if it fails
> on general errors that doesn't belong the patch's context. but it is part
> of the developing process, we can't just forget to using it. that means
> that it is mandatory to run it at some point if a developer wants his
> patches to be merged. which means the developer runs it *only* when it is
> ready to be merged. (much much less triggered jobs!)
> 
> 
> > >
> >> > There is another way though - To make the jobs do a lot less work.
> >> > Most anything has to do what what actually happens in CI resides in
> >> > the project`s automation directory now days (see [1]).  If you want to
> >> > make CI smarter so it will not do things it shouldn't be doing, all
> >> > you need to do is customize the automation scripts to be smarter and
> >> > run only the needed tests for the files that were changed by the
> >> > patch.
> >> >
> >> > [1]: http://www.ovirt.org/CI/Build_and_test_standards
> >> >
> >>
> > That's nice, but most of us are not aware of all that..
> 
> 
> > > --
> >> > Barak Korren
> >> > bkorren at redhat.com
> >> > RHEV-CI Team
> >> > _______________________________________________
> >> > Infra mailing list
> >> > Infra at ovirt.org
> >> > http://lists.ovirt.org/mailman/listinfo/infra
> >> >
> >> _______________________________________________
> >> Infra mailing list
> >> Infra at ovirt.org
> >> http://lists.ovirt.org/mailman/listinfo/infra
> >>
> >>
> >>
> >
> >
> > --
> > Eyal Edri
> > Supervisor, RHEV CI
> > EMEA ENG Virtualization R&D
> > Red Hat Israel
> >
> > phone: +972-9-7692018
> > irc: eedri (on #tlv #rhev-dev #rhev-integ)
> >
> 
> From what I'm seeing, most of the developers here don't make their patches
> drafts.. moreover,
> - personally I didn't even know that it will not trigger jobs if it is a
> draft. (and I'm not the only one)
> - sometimes I need to label my patches, therefor can't make it a draft
> 
> nowadays we are waiting for the jobs too much to finish. and the reality is
> that too much jobs shouldn't run at all- despite all of the nice things you
> guys show here..
> 
> I still think that it will be a better solution to force the developer to
> activate the tests manually (by adding a flag when pushing or even doing it
> with the jenkins client..)

That effort was called the Worflow flag, but was discarded as most of the
developers (and managers) did not like it.

It's a new flag (as CI or Verified) that you set whenever you want it to move
to the next step of the workflow, in this case, run the ci tests. Some projects
use it for requesting code reviews (most of infra projects have it), but can be
easily adapted to have as many values as needed (+1 if it needs review, +2 to
run CI jobs for example).

Another thing that would help there, is using personal/feature branches, so you
develop your patches on you branch (that does not run any CI) and at any point
you send a merge commit to the destination branch that will run the tests.
This allows:
   * Not running any ci on non-finished patches
   * For each set of related patches, only one is sent to the destination
     branch, that runs only one job (if you send all the patches, it runs for
     each)
   * You can still see the per-patch history, as the merge commit includes that
     info (as opposed to squashing it)

Those two options do not conflict either, but both of them require from the
devels to adapt their way of working.

> _______________________________________________
> Infra mailing list
> Infra at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/infra

-- 
David Caro

Red Hat S.L.
Continuous Integration Engineer - EMEA ENG Virtualization R&D

Tel.: +420 532 294 605
Email: dcaro at redhat.com
IRC: dcaro|dcaroest@{freenode|oftc|redhat}
Web: www.redhat.com
RHT Global #: 82-62605
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: not available
URL: <http://lists.ovirt.org/pipermail/infra/attachments/20151209/31ceac50/attachment.sig>