Open Stack

Steven Dake (stdake) wrote:
> Release team,
> 
> At one point Doug had indicated all projects would automatically branch
> on tagging of rc1.  I notice in git no Kolla stable/newton branch
> exists. Fwiw this is actually a good thing, because 33 patches have
> merged since rc1 relating to things that need to go into Newton,
> dramatically reducing the amount of backport work we need to do.  Part
> of this was error on my part – not validating all FFE blueprints that
> were marked Implemented were actually implemented.  One related to
> monitoring (and part of the 33 patches since rc1) was actually “Needs
> Review” rather than Implemented (as it was marked).
> 
> I don’t want Kolla to be a special snowflake wrt release processes, and
> we can live with a branch on rc1.  A branch on rc2 would be far better
> for us as we have roughly 250 bugs to triage or fix.  I leave it in the
> release team’s capable judgement to decide best on a course of action.

I'm probably the one to blame for that. Kolla follows milestones but is
trailing the release, which makes it a bit of a release snowflake. I
wasn't sure we should cut the stable branch at RC1 for such a case
(since you're still far away from final).

We should discuss what to do here (branch ASAP, branch at RC2...) on
Monday on the release channel when Doug is around.

> I would request that the expected time of branch be communicated clearly
> to us for the Newton cycle.  I have been communicating with our team
> that rc1 is where we branch.  Folks are now asking “where is the Newton
> branch of Kolla?”

FWIW Doug has been working on a spec so that projects communicate more
clearly when they want the release branch to be cut. For
milestone-driven projects it's usually clear (we branch at RC1), but for
other cases (intermediary-released, trailing) options are a bit more
open so having a way (through the openstack/releases repo) to clearly
communicate "when" will definitely help.

-- 
Thierry Carrez (ttx)