openstack team mailing list archive

[corvus@xxxxxxxxxxxx (James E. Blair)]

Johannes Erdfelt <johannes@xxxxxxxxxxx> writes:

> There appears to have been some changes to Jenkins recently. Jobs do not
> appear to be associated with Gerrit changes anymore.
>
> https://jenkins.openstack.org/job/gate-nova-python27/
>
> You can see jobs 335 and older have a link to Gerrit, but new jobs don't
> have any.
>
> This makes it harder to see what is happening in Jenkins but also
> removes the "Retrigger" link I used to use when there was a transient
> failure in a job.
>
> Was this an intended change in behavior?

Yes, see the message with subject "Parallel execution of Jenkins gate
jobs" on this list.

The short version is that we've moved most triggering of Jenkins jobs
out of the Gerrit Trigger Plugin.  The bad news about that is that we
lose the those links and the retrigger buttons.  The good news is that
we get parallel execution of gate jobs.  This means that if you approve,
say, four changes to nova, they will all test and merge at the same time
(if they all succeed).

Especially when we start gating on tempest, we are looking at
significantly longer run-times for gate tests, so we very much need the
ability to increase the rate at which we can test and merge changes.  I
think the trade off will be worth it.

You can still link Jenkins jobs to Gerrit changes though.  Near the top
of most jobs, you will see a line like:

  Triggered by: https://review.openstack.org/8228

So you can see what change triggered a Job.  Because jobs are run and
are reported more quickly (Jenkins has gone from having a _huge_ queue
of jobs to run to generally having no queue) the time between when there
have been jobs triggered and when they are reported to Gerrit should be
shorter, so there should be less need to search for a job in Jenkins.

I also miss the retrigger button.  If it is very important, we can work
on adding equivalent functionality back.  However, core members can
always retrigger a change by leaving another "Approved" vote.
Additionally, we're working on reducing transient failures -- we've just
rolled out a local pypi mirror that is built from the requirements of
the collection of projects.  Hopefully we can reduce the number of
transient failures and we won't need that functionality.

-Jim