[openstack-dev] [all] Timeframe for future elections & "Release stewards"
sean at dague.net
Wed Sep 7 16:04:54 UTC 2016
On 09/07/2016 11:43 AM, Thierry Carrez wrote:
> Hi everyone,
> As you probably know by now, starting with the Boston event in 2017, the
> Summit will happen further away from the release day and more around the
> middle of the next development cycle. You can find more info on the
> rationale for that at  and  if interested, this is not the topic
> of this email.
> One interesting side-effect is that since the timing of the election
> period (for PTL and TC positions) is defined in the TC charter
> relative to the *Summit*, it means that (unless we change this) we'll
> now run elections to renew PTL and TC positions in the middle of the
> cycle. Crazy, right ? That's what I first thought. But after discussing
> it with various people, this is not as crazy as it sounds.
> First, the current election timing is not perfect -- we change PTLs in
> the middle of the design summit prep, with old PTLs making Design Summit
> space requests that will affect their successor. It's not as if there
> was a perfect timing for doing elections.
> Second, release cycles are longer than 6 months. They actually start a
> few months before actual development starts, with discussions on next
> cycle priorities and Design Summit prep. They continue a few months
> after release, with critical stable branch backports and communication
> about landed features. So they are one year-long, overlapping cycles
> (like explained on the diagram at ). With that in mind, the PTL/TC
> election actually would happen just before the start of the start of the
> requirements-gathering pre-development phase of the next development
> cycle, which makes a lot of sense.
> Now, the main drawback of holding elections in the middle of a
> development cycle is that you don't want to introduce a discontinuity in
> leadership in that development cycle. To mitigate that, we propose the
> introduction of a new role, the "release steward", which would be
> attached to the release cycle. That person (who may or may not double as
> PTL) would be responsible for a complete release cycle on a given
> project team, from requirements gathering phase to post-release
> bugfix-backport phase. A sort of per-cycle release liaison on steroids.
> Since development cycles overlap, there would be two active release
> stewards at all times. This would help with the awkward situation where
> the PTL ends up having to think about the next cycle and prepare the
> Design Summit (or PTG) while still being knee-deep juggling with feature
> freeze exceptions, getting the current release out of the door, and
> coordinating early critical fixes stable backports. Those two jobs could
> be held by two different people.
> Now, some teams (especially those doing intermediary releases) may want
> to use the same super-human to handle everything (PTL, release steward,
> release+1 steward), and some others might use two or three humans to
> spread the load. That's up to them. But once designated by the
> newly-elected PTL, the release steward would be responsible for the full
> release cycle and would not be displaced by the next PTL 6 months later.
> One year being a long time, if a steward needs to step down, the
> currently-active PTL would appoint someone else to finish the job.
> With this new concept I think we can get the best of both worlds, and
> keep the election period as currently defined in the charter (rather
> than having to change it). The PTLs we will elect in the coming weeks
> won't be renewed before April, 2017 -- while Pike development will start
> in February.
> I know this can all be a bit confusing, so feel free to reach out to me
> with questions on this.
>  http://www.openstack.org/ptg
>  http://www.openstack.org/ptg/ptgfaq/
I think another option would be to run the PTL election early, but just
don't have the turn over happen until the master release opens up. The
current transition period is actually quite short as noted by the
comments around how design summit planning happens. Having the PTL-next
elected half way through the cycle, but having PTL current still own
landing the current release actually provides a lot more transition time.
More information about the OpenStack-dev