TripleO had PTG sessions on Monday, Tuesday, and Wednesday, followed by a
Directord+Task-Core hackfest on Thursday.
The main etherpad for the PTG is at
https://etherpad.opendev.org/p/tripleo-yoga-topics
(includes links to recordings)
Overall Summary
===============
Our sessions were well attended with good cross collaboration from external
stakeholders (storage, network, compute, etc). We had a good mix of topics,
which I felt helped maintain interest and engagement. Topics ranged from future
proposals, integrations, CI, and infrastructure. Our attendance was between
30-40 for most sessions.
The underlying theme of the week was about the proposed migration from Ansible
to Directord+task-core. Overall, I felt we accomplished clarifying the scope of
the proposed change and articulating the reasons why the change is necessary,
as well as understanding the benefits. With the heavy current investment in
Ansible specific components, we emphasized the need for migration tooling
wherever possible. No significant objections were raised about the approach.
However, we settled on the next steps being agreement in the spec, and
agreement around the end to end integration in TripleO. We continued to tie
most other topics we discussed back to this topic to make sure we were
considering the future impacts.
The hackfest on Thursday further helped to understand the proposal and
illustrated several key concepts that are presently missing in TripleO. Namely,
those of dynamic service dependency management, job result caching and
fingerprinting, and the messaging architecture. We had around 30 participants
during the hackfest.
We took a team photo on Thursday which we can be seen at
https://slagle.fedorapeople.org/TripleO-Yoga-PTG-Team-Photo.png
Thanks for the participation everyone!
Sessions
========
Monday
------
1310-1350 Directord/task-core introduction and overview
https://etherpad.opendev.org/p/tripleo-directord-task-core
The overview of Directord and task-core was presented, and we clarified the
scope of the change, which is primarily about replacing Ansible in TripleO.
As part of the change, we will need to migrate from all Ansible specific
implementations (playbooks, roles, tasks, modules, actions, etc) to pure Python
or Directord orchestrations and components. We emphasized the need for
migration tooling wherever possible. No major objections were raised, and I
felt the team understood the proposed benefits. Next steps are to finish the
spec and continue on the TripleO integration.
1405-1445 task-core task graphs and execution concepts
https://etherpad.opendev.org/p/tripleo-task-core-execution
Further illustrated the benefits and functionality gained with task-core,
particularly around the dynamic dependency ordering for tasks. Explained the
benefits for services with a lot of external deployment tasks, such as Octavia.
1500-1540 (CI) Tripleo Health
https://etherpad.opendev.org/p/tripleo-ptg-tripleo-health
Really great work by the CI team on http://ci-health.tripleo.org/ and
https://opendev.org/openstack/tripleo-ci-health-queries. We need to get the
whole team committed to maintaining the queries for the most benefit.
1540-1620 OS Migrate update/feedback session
https://etherpad.opendev.org/p/yoga-os-migrate
Update on new features added to os-migrate, and version/compatibility testing.
Tuesday
-------
1310-1350 ceph integration https://etherpad.opendev.org/p/tripleo-ceph-yoga
Plan to remove support for ceph-ansible, and add cephadm ingress with
haproxy+keepalived. Upgrade challenges around migrating from PCS to ingress. We
also need to keep ceph integration as a priority and proving ground for
migration to Directord+task-core.
1405-1445 healthchecks
https://etherpad.opendev.org/p/tripleo-healthchecks-yoga
Participation from Sean Mooney from the Nova team to present the proposed
solution in Nova. Agreement this would work well for TripleO. We need to
continue the collaboration and participate in the pending Nova spec to help
drive the healthcheck model.
1500-1540 (HA) Pacemaker brainstorming
https://etherpad.opendev.org/p/tripleo-yoga-pacemaker
General consensus is that we still need pacemaker for some tasks. But, we need
to evaluate each usage on a case by case basis and migrate services out of
pacemaker where there is opportunity.
1540-1620 FIPS Updates & Secret Management
https://etherpad.opendev.org/p/yoga-tripleo-fips
Plan to go forward with the spec for secret management in TripleO.
Wednesday
---------
1310-1350 (CI) tripleo-repos updates
https://etherpad.opendev.org/p/ci-tripleo-repos
Update on tripleo-repos and new functionality (yum config, tripleo-get-hash).
Covered ansible integration (for CI).
1405-1445 (CI) Update on CI and centos9 prep work
https://etherpad.opendev.org/p/centos-stream-9-upstream
Discussed needed CI jobs and CentOS 9 timeline.
1500-1540 Dependency clashes installing branched and non-branches repos in Zuul
jobs https://etherpad.opendev.org/p/branched-unbranched-dependency-clashes
Discussed limiting the usage of upper constraints as the overall approach.
1540-1620 RDO Releases
Agreed that providing the list of last tested hashes to RDO for releases that
TripleO doesn't support is sufficient. We reviewed the RDO/TripleO support
statement around intermediary releases that TripleO doesn't plan to maintain.
https://review.rdoproject.org/r/c/rdo-website/+/36342
Hackfest
========
The team participated in the hackfest by following through the etherpad at
https://etherpad.opendev.org/p/tripleo-directord-hackfest. Several participants
were able to get the environment up. We were able to work through various
issues as they arose.
Good discussion around UX improvements that are needed, and the apparent
benefits for different services due to the dynamic dependencies, job caching,
execution time, etc.
Also shared the POC of the end to end integration within TripleO.
--
-- James Slagle
--
