[openstack-qa] Priolity brainstorming my initial toughts

Matthew Treinish mtreinish at kortar.org
Tue Jun 18 16:02:34 UTC 2013


On Tue, Jun 18, 2013 at 07:57:48PM +0930, Christopher Yeoh wrote:
> On Mon, 17 Jun 2013 10:11:35 -0400
> Matthew Treinish <mtreinish at kortar.org> wrote:
> 
> > On Sun, Jun 16, 2013 at 04:40:19AM -0400, Attila Fazekas wrote:
> > > History of the Active VM creation performance.
> > 
> > Although, it's been a while since I tried a full run to do a side by
> > side comparison (which is difficult because testr won't pass
> > everything currently) but nose will. So I don't know if something
> > like a few tests hitting the max timeout on testr because of resource
> > contention is what is causing the longer run time. 
> 
> If not all the tests are passing with testr then that can really throw
> out the runtime. I saw similar things when I cleaned up the tests to
> all run under testr.  As soon as you have it working again I think its
> worth trying very hard to get changes gate on it. Otherwise it won't be
> long before it degrades again.

So I ran the numbers again, the testr run with concurrency=16 (just running with
the --parallel flag) on my dev box took about ~54min. Running serially with nose
only took ~47min. There were failures on testr but the same failures happen when
running with nose, so I'm guessing it's probably just something I screwed up in
my env. Also none of the failures were timeouts so it was not something that
would make testr slow down.

I also was testing my class partitioning patch for testrepository and the
numbers were more in line with what we are expecting for a parallel run.
(These also had roughly the same errors as the other runs)

I'm going to try to clean up the env and try again to get numbers without any
errors.

I've posted all the numbers that I collected here:
https://etherpad.openstack.org/testr-numbers

-Matt Treinish



More information about the openstack-qa mailing list