DIsabling nova on undercloud does not stop nova_compute container

[Rabi Mishra]

When testing nova less deploy (i.e disabling nova on the undercloud by setting enable_nova=false in undercloud.conf after the first deploy with nova), nova-compute container is left behind running after undercloud install.

3cc1b8b39db0 undercloud.ctlplane.mydomain.tld:8787/tripleomaster/centos-binary-nova-compute-ironic:current-tripleo kolla_start 52 minutes ago Up 52 minutes ago nova_compute

and the wrapper systemd service (tripleo_nova_compute) is still running.

There is only one startup config left
[CentOS-8.2 - root@undercloud ~]# ls /var/lib/tripleo-config/container-startup-config/step_5/
mistral_db_populate.json

And it seems we've some logic in needs_delete filter not to clean orphaned containers when there is only one config or no config.

https://github.com/openstack/tripleo-ansible/blob/master/tripleo_ansible/ansible_plugins/filter/helpers.py#L119-L126

So, there seems to be 2 issues:

1. When there is 1 config for a step, orphaned containers are not cleaned.
2. When there is no config for a step (ex. when both nova and mistral are disabled and there is no startup config for step5), it seems contents of
/var/lib/tripleo-config/container-startup-config/step_5/ are not cleaned. Hence none of the containers are stopped.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/748598

[Alex Schultz (alex-schultz)]

Historically if you needed to cleanup a service after disabling it, you would need to provide a disabled service definition that did the cleanup explicitly. I don't think we should automagically clean things up if not provided as some folks do this intentionally (for whatever reason)

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/748598
Committed: https://git.openstack.org/cgit/openstack/tripleo-ansible/commit/?id=1b93165288341659c956cdf71fa1903d62d61d7e
Submitter: Zuul
Branch: master

commit 1b93165288341659c956cdf71fa1903d62d61d7e
Author: Rabi Mishra <email address hidden>
Date: Fri Aug 28 14:15:06 2020 +0530

    Expose tripleo_container_manage_clean_orphans var

    This exposes new var for tripleo_container_manage. Also,
    removes the check to force set clean_orphans=True in
    needs_delete filter, when there is one or no item in
    the startup config.

    There is a possibility when disabling services, there would
    be one or zero container_startup_configs for a step.

    Partial-Bug: #1893335
    Change-Id: I9d08168015487c48d8b380a9575ba236b7fb7d0d

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/ussuri
Review: https://review.opendev.org/749176

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/train
Review: https://review.opendev.org/749179

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/749179
Committed: https://git.openstack.org/cgit/openstack/tripleo-ansible/commit/?id=a1f78497daad5c5bf22a5828118da724a3f1af66
Submitter: Zuul
Branch: stable/train

commit a1f78497daad5c5bf22a5828118da724a3f1af66
Author: Rabi Mishra <email address hidden>
Date: Fri Aug 28 14:15:06 2020 +0530

    Expose tripleo_container_manage_clean_orphans var

    This exposes new var for tripleo_container_manage. Also,
    removes the check to force set clean_orphans=True in
    needs_delete filter, when there is one or no item in
    the startup config.

    There is a possibility when disabling services, there would
    be one or zero container_startup_configs for a step.

    Partial-Bug: #1893335
    Change-Id: I9d08168015487c48d8b380a9575ba236b7fb7d0d
    (cherry picked from commit 1b93165288341659c956cdf71fa1903d62d61d7e)

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/749176
Committed: https://git.openstack.org/cgit/openstack/tripleo-ansible/commit/?id=17e6d8c82be168bf6b059ee32e37387659e31f8c
Submitter: Zuul
Branch: stable/ussuri

commit 17e6d8c82be168bf6b059ee32e37387659e31f8c
Author: Rabi Mishra <email address hidden>
Date: Fri Aug 28 14:15:06 2020 +0530

    Expose tripleo_container_manage_clean_orphans var

    This exposes new var for tripleo_container_manage. Also,
    removes the check to force set clean_orphans=True in
    needs_delete filter, when there is one or no item in
    the startup config.

    There is a possibility when disabling services, there would
    be one or zero container_startup_configs for a step.

    Partial-Bug: #1893335
    Change-Id: I9d08168015487c48d8b380a9575ba236b7fb7d0d
    (cherry picked from commit 1b93165288341659c956cdf71fa1903d62d61d7e)