Open Stack

On 10/03/2017 01:06 PM, Jeremy Stanley wrote:
> On 2017-10-03 12:00:44 -0500 (-0500), Matt Riedemann wrote:
>> On 10/3/2017 11:40 AM, Monty Taylor wrote:
>>> Our nodepool quota will be allocated 80% to Zuul v2, and 20% to ZUul v3. 
>>> This will slow v2 down slightly, but allow us to continue to exercise v3
>>> enough to find problems.
>>>
>>> Zuul v2 and v3 can not both gate a project or set of projects.  In
>>> general, Zuul v2 will be gating all projects, except the few projects that
>>> are specifically v3-only: zuul-jobs, openstack-zuul-jobs, project-config,
>>> and zuul itself.
>>
>> So if v3 is in check and periodic, and will be restarted and offline at
>> times, doesn't that mean we could have patches waiting for an extended
>> period of time on v3 results when the v2 jobs are long done? Or do the v3
>> jobs just timeout and being non-voting shouldn't impact the overall score?
> 
> Not at all, since v2 will also be running check jobs itself and it
> only pays attention to its own results for that puropose. So
> basically while we're rolled back to v2 you'll see both "Jenkins"
> (v2 account) and "Zuul" (v3 account) reporting check results most of
> the time but you can consider the v3 results merely an advisory
> indicator of whether your v3 jobs are working correctly.

Right, so zuul v3 will effectively be treated like 3rd party CI during
the transition. Sounds good.

The 80 / 20 split seems very reasonable, and a good way to let teams get
back to work while letting the v3 effort make forward progress with real
load to smoke out the issues.

Thanks for flipping over to this model.

	-Sean

-- 
Sean Dague
http://dague.net