10:19 a.m.
On 09. 08. 19 9:45, Sebastian Laskawiec wrote:
On Thu, Aug 8, 2019 at 5:32 PM Bruno Oliveira <bruno(a)abstractj.org
<mailto:bruno@abstractj.org>> wrote:
I believe Hynek's changes help us a lot with the PRs and solves our
problem for now. Yesterday after merging Stan's PR which only touches
into non-Java files, the build from master started to fail for the
cross-dc Job.
Cross DC builds as Authz builds run conditionally against
pull-request,
not against master. Here is my suggestion:
1. For these builds that have been consistently failing. We apply the
same conditions that we have for pull-request for master. Into other
words, if I didn't touch into cross-dc stuff, I don't run any tests
related to this when my changes are merged.
Is there any benefit on implementing this conditional runs? Perhaps we
should always run the whole test suite. This way we could discover
unstable tests faster and hopefully get them fixed.
For cross-dc tests, the problem is that they are not so stable. Running
it always will decrease travis stability and in most cases, running them
is not really necessarily. But as you pointed, there is possible issue
that we won't discover issues early and cross-dc tests can become broken
even with changes, which doesn't seem to be related to cross-dc (but
that happen only once during last 1.5 years since we have cross-dc
tests). So there are both pros and cons, but I vote to rather keep
cross-dc tests in travis conditional.
For authz tests, I am not sure. If they are stable, we can run them
always in theory. But I am not sure if they are really stable?
Marek
2. We have a Daily Job running on Travis against master. Into our
scripts we identify the event using TRAVIS_EVENT_TYPE and if that's a
"cron", we run all the tests: cross-dc, authz and others.
If that makes sense and we have some agreement about this. I can
give it
a try.
On 2019-08-05, Marek Posolda wrote:
> On 05. 08. 19 19:13, Stan Silvert wrote:
> > On 8/5/2019 3:43 AM, Hynek Mlnarik wrote:
> > > The PRs are blocked by two issues:
> > > 1) Adapter tests sometimes take too much time: This should be
> > > addressed by (now merged) [1]
> > > 2) OpenShiftTokenReviewEndpointTest being unstable,
addressed by [2]
> > >
> > > IIRC, we had issues with Travis caches [3, 4], but if there
is time it
> > > might be worth checking whether the issues persist.
> > >
> > > Once [2] is merged, Travis should be stable again for
rebased PRs.
> > Do we know when this is going to happen?
>
> PR is approved by Hynek and me, but travis didn't pass for the
PR due the
> adapter tests exceed the 50 minutes limit. I am retrying one
more time.
>
> I agree with what Bruno's suggested to split adapter tests into
more groups.
> Or fix adapter tests to not require realm re-import after each
test as that
> take significant amount of time. See [1]. But that will probably
require
> some more changes and may be probably few days of work...
>
> [1]
https://github.com/keycloak/keycloak/blob/master/testsuite/integration-ar...
>
> Marek
>
>
> >
> > > --Hynek
> > >
> > > [1] https://github.com/keycloak/keycloak/pull/6209
> > > [2] https://github.com/keycloak/keycloak/pull/6207
> > > [3] https://issues.jboss.org/browse/KEYCLOAK-5124
> > > [4] https://issues.jboss.org/browse/KEYCLOAK-7285
> > >
> > > On Sat, Aug 3, 2019 at 12:52 PM Bruno Oliveira
<bruno(a)abstractj.org <mailto:bruno@abstractj.org>
> > > <mailto:bruno@abstractj.org
<mailto:bruno@abstractj.org>>>
wrote:
> > >
> > > Before answer this e-mail I was doing some experiments
and there are
> > > some things that I identified we can do about this.
> > >
> > > 1. Enable cache again on Travis
> > >
> > > That won't fix the issue, but may help
> > >
> > > 2. Split the adapter tests
> > >
> > > I did some test with the following PR:
> > > https://github.com/keycloak/keycloak/pull/6221/files
> > >
> > > Here's the output:
> > > https://travis-ci.org/keycloak/keycloak/builds/567252894
> > >
> > > 3. Review the tests we run on Travis
> > >
> > > I'm not so sure if everything that we have running on
Travis today is
> > > really required.
> > >
> > >
> > > For now if others agree on item 2, I can polish my PR
and work to
> > > get it merged.
> > >
> > > On 2019-08-02, Stan Silvert wrote:
> > > > The adapter tests are timing out on all new PR's.
> > > >
> > > > Hynek, I'm told that you might be able to help?
> > > >
> > > > Stan
> > > >
> > > > _______________________________________________
> > > > keycloak-dev mailing list
> > > > keycloak-dev(a)lists.jboss.org
<mailto:keycloak-dev@lists.jboss.org>
<mailto:keycloak-dev@lists.jboss.org
<mailto:keycloak-dev@lists.jboss.org>>
> > > > https://lists.jboss.org/mailman/listinfo/keycloak-dev
> > >
> > > --
> > >
> > > abstractj
> > >
> > _______________________________________________
> > keycloak-dev mailing list
> > keycloak-dev(a)lists.jboss.org <mailto:keycloak-dev@lists.jboss.org>
> > https://lists.jboss.org/mailman/listinfo/keycloak-dev
>
>
--
abstractj
_______________________________________________
keycloak-dev mailing list
keycloak-dev(a)lists.jboss.org <mailto:keycloak-dev@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/keycloak-dev