[openstack-dev] [glance] Flakey functional test (Was: [Openstack-stable-maint] Stable check of openstack/glance failed)

Ian Cordasco sigmavirus24 at gmail.com
Wed Nov 16 14:02:51 UTC 2016


-----Original Message-----
From: tomislav.sukser at telekom.de <tomislav.sukser at telekom.de>
Reply: OpenStack Development Mailing List (not for usage questions) <openstack-dev at lists.openstack.org>
Date: November 16, 2016 at 07:46:18
To: openstack-dev at lists.openstack.org <openstack-dev at lists.openstack.org>
Subject:  Re: [openstack-dev] [glance] Flakey functional test (Was: [Openstack-stable-maint] Stable check of openstack/glance failed)

> > Hi Glance team members,
> >
> > Over the weekend one of our stable periodic jobs failed. It failed on
> > a test (glance.tests.functional.test_reload.TestReload.test_reload)
> > that I've seen fail a couple times previously. I've created a bug for
> > this: https://bugs.launchpad.net/glance/+bug/1641670 And I'm hoping
> > someone will be able to reproduce it and fix it.
> >
> > I suspect, this is a matter of resource contention in the donated VMs
> > that our CI uses, but I can't be certain.
> >
> > The stable team would greatly appreciate help diagnosing and fixing this issue.
> Hi,
> Even I was unable to reproduce the issue, I found something slightly odd.
> The reloaded log file is named etcnew.log and placed in the same directory
> as api.log, registry.log and others. Look like this is due to the name
> construction (in glance/tests/functional/test_reload.py:245):
> conf_dir = os.path.join(self.test_dir, 'etc')
> log_file = conf_dir + 'new.log'
> Although this might be nice to fix, it doesn't look like a problem to me.
> Since there's nothing pointing to any problems here, I would just ask is
> it possible that log file is not created if there's nothing to log?

I don't think that's possible. We start glance's services (when under test) with debug=True and verbose=True which means the config options that are collected and then parsed should be logged to the file, at least.

> (and only if that is the case - I would suggest adding some dummy request
> Which would result in log entry; if not - please ignore this idea)

Thanks for the help though, Tomislav! Did you look for the hash seed that tox was using and try running the tests that way? I've been swamped with other responsibilities this week so I haven't had time to investigate this myself.

Ian Cordasco

More information about the OpenStack-dev mailing list