Open Stack

On Mon, Jan 04, 2016 at 11:31:40AM +0100, Julien Danjou wrote:
> On Mon, Jan 04 2016, Steven Hardy wrote:
> 
> Hi,
> 
> > Firstly, I'm very sorry for the breakage here, and I agree that in general
> > a quick-revert is the best policy when something like this happens.
> 
> No problem Steven, shit happens.
> 
> It'd be even better if Heat'd move to a devstack plugin to limit this
> kind of high level latency. Just by curiosity, is there a plan for that?

Yes, it was discussed in a recent Heat meeting, and IIRC pas-ha was going
to work on it, I agree it's a good idea and should enable more responsive
updates/fixes to our devstack config.

> > I'm a little unclear how this occurred tho, since I had a clear CI run on
> > this patch:
> >
> > https://review.openstack.org/#/c/256315/
> 
> I think the dsvm tests are not the same that we run on telemetry side:
> we run it with the Gnocchi backend, whereas you're likely using the
> (old) Ceilometer backend. And that's the Gnocchi backend that has been
> broken by the original devstack change from what I saw.
> 
> Maybe Heat should also run this job, WDYT?

We'll have to look into that - I was assuming all our jobs should move to
using Gnocchi by default now?

Rabi has acutally pointed out that we'd already skipped the heat alarm test
due to a previous issue (doh!) which is probably why my Depends-On didn't
catch this regression, so at least that is now understood :)

Steve