[TripleO] centos9 jobs only for master, centos 8 & 9 for wallaby

Jesse Pretorius jesse at odyssey4.me
Tue Mar 1 14:57:48 UTC 2022

Hi folks, response in-line.

> On 24 Feb 2022, at 15:11, Marios Andreou <marios at redhat.com> wrote:
> Hello TripleO o/
> During the last period the tripleo-ci team has been working on getting
> tripleo CI jobs to run centos-9.
> On master branches across all TripleO repos we now run C9 exclusively
> (since [1]).
> On Wallaby branches, after the work at [2] (in particular after
> https://review.opendev.org/c/openstack/tripleo-ci/+/830132 merges) we
> will run both C8 and C9 jobs. I started an etherpad at [3] thinking
> this would be a PTG topic, but I think we'll want to have this
> conversation way before then. There is a tripleo-heat-templates
> example in the etherpad that shows what we can expect from the job
> runs (note the exact jobs that run will vary per repo and even between
> changes depending on the files touched - but generally there will be
> duplication between 8 and 9 jobs).
> The current proposal is that we keep a minimal subset on C8 Wallaby,
> with C9 wallaby having the full set of jobs.
> Two factors that will affect our decisions are (there may be more?)
>  i) Upgrades requirements - are we supporting upgrading to Wallaby on
> 8? For example, the coming undercloud-upgrade-ffu job will be train 8
> to wallaby 8. In which case we definitely need to keep at least some
> subset of 8 jobs (and can't entertain the removal of 8 from Wallaby
> completely).

I think trying to keep up with the moving target of 8-stream would be counterproductive given that we are pinned downstream to RHEL 8.4 for Train. We do have a pending patch to merge a job for 8-Stream/Train->8-Stream Wallaby for the Undercloud only, which would be nice to keep but if it becomes troublesome then I’d suggest we remove it. This job adds some value to us (and has added quite a bit already in preparing it and making it pass), but if 8-stream starts making it fail, or it starts failing due to database migrations then we’ll have to shift that job downstream and deal with the consequences of only ever finding upgrade bugs post-merge.

It is well known that a Fast-Forward Upgrade (FFU) of OpenStack services is not supported upstream, so we can not expect any upstream jobs to reliability support this process. If/when there is broader support for enabling FFU support officially in OpenStack services, then we can reconsider this position.

>  ii) Are we importing from Wallaby 8 or Wallaby 9? Currently it is 8
> but this will soon switch.
> For the wallaby c8 'subset of jobs' e.g. multinode, vanilla standalone
> (no scenarios? some subset of them?), undercloud-ffu, minor update.
> This is just to start the conversation so please reply if you have
> thoughts or comments about any of the above.
> We are planning to discuss this in the coming tripleo-ci community
> call this coming Tuesday at 1330 UTC - meeting link at [4] so please
> join us if you can and would like to participate,
> regards, marios
> [1] https://review.opendev.org/q/topic:c8_teardown_master
> [2] https://review.opendev.org/q/topic:c9_wallaby_gates
> [3] https://etherpad.opendev.org/p/tripleoci-wallaby-centos-8-9
> [4] https://meet.google.com/bqx-xwht-wky

More information about the openstack-discuss mailing list