[openstack-dev] [nova] gerrit based statistics

Daniel P. Berrange berrange at redhat.com
Thu Oct 9 09:23:27 UTC 2014


On Wed, Oct 08, 2014 at 04:30:16PM -0700, Joe Gordon wrote:
> Recently there has been a lot of discussion around the development growing
> pains in nova. Instead of guessing about how bad some of the issues are, I
> tried to answer a few questions that may help us better understand the
> issues.

What time period did you calculate these stats over ? One whole dev
cycle ?

> Q: How many revisions does it take to merge a patch?
> 
> Average: 6.76 revisions
> median: 4.0 revisions
> 
> 
> Q: How many rechecks/verifies does it take to merge a patch (ignoring
> rechecks where the same job failed before and after)?
> 
> Average: 0.749 rechecks per patch revision
> median: 0.4285  rechecks per patch revision

These feel really low to me, but I guess that is a consequence of
aggregating over a long time period.

> For comparison here are the same results for tempest, which has a lot more
> gating tests:
> 
> Average: 1.01591525738
> median: 0.6
> 
> 
> Q: How long does it take for a patch to get approved?
> 
> Average: 28 days
> median: 11 days
> 
> 
> Q: How long does it take for a patch to get approved that touches
> 'nova/virt/'?
> 
> Average: 34 days
> median: 18 days
> 
> 
> When looking at these numbers two things stick out out:
> 
> * We successfully use recheck an awful lot. More then I expected
> * Patches that touch 'nova/virt' take about 20% more time to land or about
> 6 days. While that is definitely a difference, its smaller then I expected
> 
> 
> Dataset: last 800 patches in nova
> Code: https://github.com/jogo/gerrit-fun

Ah ok, so I think what would be more informative is to caculate the
stats on weekly blocks and graph it over 2 cycles. There are certainly
times when the problems are much worse than others and a long term
average will mask that to a large extent. It is the periods when it is
particularly bad that are leading to the high frustration amongst
contributors IMHO, so we must look at the worst cases.

For example look at the stats I produced for patch review comments per
core team reviewer:

  https://www.berrange.com/~dan/novacodeandspecsreviewrate.png

The gap betweeen good weeks and bad weeks is enourmous.


Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|



More information about the OpenStack-dev mailing list