I am running some more tests now with the NetVirt CSIT that look promising
so it might not be a blocker for NetVirt. I am running a few more
iterations now to know better.

I had reduced the number of test suites so that we could capture just
clustering issues. Doing so did add a bug in the test code that was causing
some extra failures. I have that fixed. If the next runs show we are back
down to just a few clustering bugs then we don't need a blocker from the
NetVirt side. If lucky the remaining issues are what is in this
openflowplugin bug here.

For reference the last run [1] is looking better. It has a different test
code bug in so ignore those, but if you can check the karaf.logs and see if
you see any clustering issues. I don't think you will since the killing of
the ODL nodes is broken in this job. 21 and 21 should have that fixed.

[1]
https://logs.opendaylight.org/releng/jenkins092/netvirt-csit-3node-openstack-ocata-gate-stateful-carbon/20/

On Thu, Aug 24, 2017 at 7:22 PM, Robert Varga <[email protected]> wrote:

> On 24/08/17 22:07, [email protected] wrote:
> > *Comment # 14 <https://bugs.opendaylight.org/show_bug.cgi?id=9006#c14>
> > on bug 9006 <https://bugs.opendaylight.org/show_bug.cgi?id=9006> from
> > Luis Gomez <mailto:[email protected]> *
> >
> > OK, I think as next step I can try to see if this reproduces outside CI.
>
> +infrastructure
>
> Guys, we are dealing with an issue which was first reported on
> 8/17/2017, is blocking Carbon SR2 (due to netvirt CSIT failing) and can
> either be an infra or code problem.
>
> Suspect trigger is the fix for
> https://bugs.opendaylight.org/show_bug.cgi?id=8941 (merged of
> 8/12/2017), which is a Carbon -> Carbon SR1 memory leak regression. If
> that is the case, we need to identify and fix it, as a revert is not
> really an option.
>
> Carbon/Nitrogen are synced up w.r.t. CDS, so this also impacts Nitrogen
> (where it is a Carbon -> Nitrogen regression).
>
> Can you check with RS if there are any issues and/or the public cloud is
> experiencing issues?
>
> Given that inter-node network stability is in question, can we get a
> limited-use set of slaves in the private cloud? Whatever is needed for
> netvirt CSIT is sufficient, and we only need to spin it up when we need
> a really predictable environment... Should I file a helpdesk ticket?
>
> Thanks,
> Robert
>
>
> _______________________________________________
> infrastructure mailing list
> [email protected]
> https://lists.opendaylight.org/mailman/listinfo/infrastructure
>
>
_______________________________________________
infrastructure mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/infrastructure

Reply via email to