I also believe we can find ways to make post-merge / periodic checks useful.
We need to do that to keep the gate to a sane scale.

On 24 October 2014 17:33, Thierry Carrez <thie...@openstack.org> wrote:
> Ihar Hrachyshka wrote:
>> Another question to solve is how we disseminate state of those jobs.
>> Do we create a separate mailing list for that? Obviously we should not
>> reuse -dev one, and it's overkill to create one mailing list per
>> interest group.
> Should we explore other avenues than email for this ? If we plan to do
> opt-in anyway, would some status website/RSS not work better ?


> The ideal system imho would be a status website where we could see
> failures and close them as "handled" so that everyone knows that a past
> FAIL result has already been fixed. That could help avoid duplication of
> painful debugging work.


Publicizing the test results better, and to the interested audience
will help a lot.
Same as keep a track record of fixed issues and solutions.

Tracking result history at test level (using subunit2sql), build and
analyze trends would be a great tool to identify and troubleshoot

Also be beneficial IMO would be extracting whatever information can be
gather automatically from the test results.
Rather than saying "job X failed" we could have tools that allow us to
tell "test X started failing in a specific time range, and this is the
list of sha1s that have been merged around that time."

We will also discuss about this topic at Paris in the QA track.

Andrea Frittoli (andreaf)

OpenStack-dev mailing list

Reply via email to