[Pulp-dev] New Year's resolution: ending Pulp-Qpid deadlocks.

Alan Conway aconway at redhat.com
Fri Jan 6 22:43:44 UTC 2017

On Wed, 2017-01-04 at 10:30 -0500, Sean Myers wrote:
> On 01/03/2017 05:19 PM, Alan Conway wrote:
> > I want to put an end to the pulp-qpid deadlock, AKA
> > 
> > https://bugzilla.redhat.com/show_bug.cgi?id=1377195
> > https://issues.apache.org/jira/browse/QPID-7317
> > 
> > but I am still having trouble reproducing it in a controlled
> > environment. I'm launching an appeal to anyone with ideas about how
> > to
> > reproduce this problem semi-reliably so we can fix it once and for
> > all.
> We normally see this happen in our jenkins automation, so naturally
> today
> is the day that our rhel7 images break. In my experience,there's no
> single
> reproducer other than "doing a lot of stuff with pulp"; it doesn't
> reliably
> fail on a certain test or running for certain amount of time.
> I'm working on getting the el7 images working again, and once they're
> back
> up we might at the very least be able to set up some rudimentary
> monitoring
> that can send alerts when we get a probable deadlocked jenkins node.

That might help. Meantime I'm setting up a VM rhel7 environment that I
will dedicate to running tests where this seems to crop up most often
on an intensive basis. It's nearly ready to go, the learning curve for
setting up VMs etc. was steeper than I expected.

More information about the Pulp-dev mailing list