15:00:43 <noonedeadpunk> #startmeeting openstack_ansible_meeting
15:00:43 <opendevmeet> Meeting started Tue Aug  2 15:00:43 2022 UTC and is due to finish in 60 minutes.  The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:43 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:43 <opendevmeet> The meeting name has been set to 'openstack_ansible_meeting'
15:00:50 <noonedeadpunk> #topic rollcall
15:00:53 <noonedeadpunk> o/
15:00:55 <damiandabrowski> hi!
15:03:07 <jrosser_> o/ hello
15:03:53 <mgariepy> o/ hi !
15:04:28 <noonedeadpunk> #topic office hours
15:04:40 <noonedeadpunk> so seems, that rocky jobs have not returned back
15:05:04 <NeilHanlon> o/
15:05:16 <noonedeadpunk> or well, it's only for Yoga I guess. So I likely is wrong here :)
15:05:35 <NeilHanlon> they should be resolved today. ianw was working on a dib release, I believe
15:05:52 <noonedeadpunk> aha, would that involve r9?
15:06:25 <NeilHanlon> no, just fixing the r8 ones. but my change for r9 should be able to come pretty quickly, I hope
15:07:01 <noonedeadpunk> ok, great!
15:07:42 <noonedeadpunk> another thing we were reported recently, is that we have duplicated records in hosts file, if deploy_host == controller as an example
15:08:26 <noonedeadpunk> basically we likely should be somehow smarter in https://opendev.org/openstack/openstack-ansible-openstack_hosts/src/branch/master/defaults/main.yml#L40
15:09:26 <noonedeadpunk> not really sure how to check for that. as an inventory this host won't be == to 'localhost'
15:09:42 <noonedeadpunk> *in inventory
15:09:46 <NeilHanlon> hrm..
15:10:17 <noonedeadpunk> can't say it brings a lot of issues, as records are identical, but that's confusing at very least
15:11:12 <NeilHanlon> maybe instead of the join, we could filter the list or unique it
15:12:02 <mrf> Openstack installed ! :D tomorrow will learn how to add more nodes to a deployment thank so much for your work and help!
15:12:12 <noonedeadpunk> nah, it's a bit different. So this variable jsut controlls if to run this task or not https://opendev.org/openstack/openstack-ansible-openstack_hosts/src/branch/master/tasks/openstack_update_hosts_file.yml#L48-L57
15:12:13 <NeilHanlon> _etc_hosts_content | unique | join('\n') or so https://opendev.org/openstack/openstack-ansible-openstack_hosts/src/commit/f26fbe4c6aef11cc17c4a9e7be37a7b89ae74326/tasks/openstack_update_hosts_file.yml#L51
15:12:29 <noonedeadpunk> So we jsut place same content twice
15:12:38 <noonedeadpunk> under different blocks
15:12:58 <NeilHanlon> yeah
15:13:59 <noonedeadpunk> on L40 and L48. and it's not that content is wrong, we should just somehow avoid doing that
15:14:37 <NeilHanlon> oh. my bad, I understand the problem now
15:16:04 <noonedeadpunk> anyway, it doesn't hurt much and there's a control, so maybe we should just document that better
15:16:33 <NeilHanlon> only thing I can think of doing is looping over the list and doing a lineinfile to ensure the lines are inserted once and only once; but that's computationally annoying :)
15:17:46 <noonedeadpunk> well, the bigger problem with that approach is how to cleanup obsolete lines
15:17:59 <noonedeadpunk> in case compute node is dropped or container re-created with different IP
15:18:43 <noonedeadpunk> and you can not really clean-up everything not in your list, as ppl might have some custom records there
15:19:54 <noonedeadpunk> Yestarday jrosser_ also spotted weird behaviour in our dynamic_inventory script. While we should patch it right now, this brings me to think what should be our futher plan for this>
15:20:42 <noonedeadpunk> as dynamic_inventories are being deprecated. One way would be moving it to inventory plugin. But maybe we should jsut have a static inventory, that's being generated/managed with help of inventory-manage script
15:20:57 <NeilHanlon> Simple way would be not supporting the controller == deploy host scenario, but that's not ideal :D
15:21:19 <jrosser_> i don't think that the odd inventory behaviour will affect many deployments, but it's just wrong and extremely confusing
15:21:33 <jrosser_> though it does make a really wierd deployment of ironic on LXC
15:21:53 <noonedeadpunk> well, it affects all deployments, but unlikely you will notice that unless do deep dive into it
15:22:35 <jrosser_> i think we now understand what is happening, but not necessarily yet how to fix it
15:22:40 <noonedeadpunk> so it's more of a corner-case usage I would say
15:22:56 <noonedeadpunk> yeah
15:27:44 <noonedeadpunk> regarding publishing config_template to galaxy - my original didn't work, as you can not read secrets that are stored in different project except it is project-config.
15:28:33 <noonedeadpunk> I tried to push jobs to project-config, and agreed with keys-keeper to share secret with us, but this job got -1 from infra stuff with suggestion to move it partially to zuul-jobs.
15:29:18 <noonedeadpunk> I didn't have time to follow-up on that, but getting it to zuul-jobs sounds weird to me, mainly because of the code and I bet I will get -2 there...
15:29:48 <noonedeadpunk> at the same time I do not want to copy-paste jobs code, but well... this is other way around
15:31:17 <noonedeadpunk> We've also landed almost everything to stable/xena to do bump and next release. This in turn would unblock Yoga upgrade jobs.
15:31:42 <jrosser_> we have done another X->Y upgrade today
15:31:54 <jrosser_> andrewbonney may have a few small things to fix
15:32:09 <jrosser_> i think we have some docs errors around upgrades
15:32:23 <noonedeadpunk> yeah, we don't mention octavia at very least
15:32:56 <noonedeadpunk> in terms of moving certs to be usable by pki role
15:38:27 <noonedeadpunk> I don't think I have anything extra to add though
15:44:07 <jrosser_> no - do we have any new bugs?
15:48:17 <noonedeadpunk> I don't think we do?
15:48:33 <jrosser_> i've not looked :)
15:48:46 * jrosser_ might have fixed the inventory
15:48:56 <jrosser_> ^ extremely unlikley, but possible
15:49:03 <noonedeadpunk> we have not closed some recent ones though, but we didn't have new since few weeks I guess
15:49:11 <noonedeadpunk> and last one is assigned to damiandabrowski :)
15:49:25 <noonedeadpunk> oh
15:51:54 <damiandabrowski> yeah sorry, will try to come back to it during this week :/
15:55:05 <opendevreview> Dmitriy Rabotyagov proposed openstack/openstack-ansible master: Do not create {hostname}-host_containers group as child of other groups  https://review.opendev.org/c/openstack/openstack-ansible/+/851764
15:57:29 <noonedeadpunk> Just did some extra cleanup to the removal of host_containers - not fixing state unfortunately
15:58:58 <noonedeadpunk> #endmeeting