Open Stack

Hi folks,

In the long-term effort of simplification and convergence to Ansible, the
replacement of Paunch has passed a big milestone where it has proven to be
stable enough to be the default.
A lot of testing has happened, including updates and upgrades; all positive
so far.
The role itself is documented here:
https://docs.openstack.org/tripleo-ansible/latest/roles/role-tripleo-container-manage.html
It has functional testing (Molecule), testing things like container config
updates, idempotency, and many more use cases.
You can take a look:
https://opendev.org/openstack/tripleo-ansible/src/branch/master/tripleo_ansible/roles/tripleo_container_manage/molecule/default/playbook.yml
The role itself is using custom filters; which are all unit tested. Some
modules were created as well. More testing has to happen on these modules
but they have proven to work and even improve the operator experience to
cover more scenarios and improve the logging and debugging process.
The role doesn't support Docker, and won't at this point; it doesn't fit
with our roadmap.

With that patch, any new deployment (not using Docker) will now use the new
role and not Paunch anymore:
https://review.opendev.org/#/c/700738/
Which means: standalone, undercloud, overcloud.

If you already have a running deployment, you can either run a minor update
which will update THT and roll your cloud; or you can also set
EnablePaunch: False and the containers will be restarted with the new
config.

If for some reason, you need to disable it, here's how:
- undercloud.conf: undercloud_enable_paunch = false
- standalone/overcloud: EnablePaunch: False

It is not supported to roll back to a Paunch deployment once your
containers are managed by Ansible; if you try it, an error will raise and
the deployment/update/upgrade will stop.


If you find any problem, bug or have any feedback for improvement, please
let me know and we'll make it better.
During the following weeks, we'll watch for performances and see if it
degraded or if it's better. Paunch's replacement shouldn't make it faster
because Ansible has a lot of overhead; however we want to make sure that it
doesn't cause issues at scale. Which is why we're going to observe that now
and do more testing at scale now it's the default.

I want to thank all the reviewers involved in that effort, and specially
Sagi, Kevin and Alex for their great feedback.
Thanks,
-- 
Emilien Macchi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20200221/29147ad5/attachment-0001.html>