Improve rabbitmq reset playbook matching #1103
Conversation
Is this always true? I reckon a container could get into a failed state due to rabbit issues, it might be better to match against the description as this is the last field and always matches the kolla container name. |
Alex-Welsh
force-pushed
the
fix-rabbit-reset
branch
from
89b65e2 to
e220685
Compare
Feels a bit wrong to match against the description rather than the ID but fair point |
| @@ -68,4 +68,4 @@ | |||
| # The following services use RabbitMQ. | |||
| - name: Restart OpenStack services | |||
| shell: >- | |||
| systemctl -a | egrep '(barbican|blazar|cinder|cloudkitty|designate|heat|ironic|keystone|magnum|manila|neutron|nova|octavia)' | awk '{ print $1 }' | xargs systemctl restart | |||
| systemctl list-units --type=service | egrep '(barbican|blazar|cinder|cloudkitty|designate|heat|ironic|keystone|magnum|manila|neutron|nova|octavia)' | egrep -v ceph | awk '{ print $NF }' | xargs systemctl restart | |||
There was a problem hiding this comment.
Are you sure about $NF in your awk here?
There was a problem hiding this comment.
I see this is to match the description. A bit dangerous if this changes in kolla-ansible.
There was a problem hiding this comment.
New approach, splitting every word out to a new line and selecting unique entries with better regex matching (must start with kolla-, end with -container.service
There was a problem hiding this comment.
Will this now pick up each container twice, from the name and description? Could there be any issues from calling the restart twice in quick succession?
There was a problem hiding this comment.
systemctl list-units supports passing a pattern in, can we do that instead of the grep? https://www.freedesktop.org/software/systemd/man/latest/systemctl.html#list-units%20PATTERN%E2%80%A6
There was a problem hiding this comment.
I added | uniq to stop it duplicating lines.
I don't think the systemctl pattern really helps. Grep is more versatile here
Alex-Welsh
force-pushed
the
fix-rabbit-reset
branch
from
e220685 to
e82e0fe
Compare
Alex-Welsh
force-pushed
the
fix-rabbit-reset
branch
from
e82e0fe to
22a35a8
Compare
|
@Alex-Welsh It's always good to check if a PR attempting to fix your issue is not already opened :) |
|
Oh, I see that you've seen that one (as you've upvoted Michal's suggestion) :) |
I can barely remember what I had for breakfast this morning, let alone anything that happened in May 😆. I'm happy to close this one and continue with yours if you update it |
|
@Alex-Welsh actually, it's my bad! I fixed matching aswell and never pushed to that branch I pasted earler! That's why I found this redundant. Great job and thanks! |
|
https://github.com/stackhpc/stackhpc-kayobe-config/blob/stackhpc/2023.1/etc/kayobe/ansible/stop-openstack-services.yml Just realised we might want to change this too, as it will also be matching against the dot on failed services. Less impactful as stopping a failed service probably won't do much, but still seems worth doing. |
|
Closing in favor of #1073 |
Two issues were picked up during the Sussex deployment. The original command matched against the correct lines but also these:
The first issue is that it was matching against ceph containers which happen to have
manilain the nameThe second was that it was matching against failed units, which show up in the list like this:
So it's matching against the dot
Stopped containers don't need to be stopped and restarted so we should only match against running units