15:00:04 #startmeeting openstack_ansible_meeting 15:00:04 Meeting started Tue Feb 28 15:00:04 2023 UTC and is due to finish in 60 minutes. The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:04 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:04 The meeting name has been set to 'openstack_ansible_meeting' 15:00:10 #topic rollcall 15:00:11 o/ 15:00:49 o/ hello 15:02:51 #topic office hours 15:04:16 I don't have much this time. I've started looking into quorum queues and by far found that plugins repo does not test actual code. Repo is in required-projects but likely we don't install collection from zuul path for some reason 15:04:45 jrosser: I know you've proposed patch to change approach there a bit but it failed as well - I didn't have time to check what went wrong there 15:05:23 will have some deeper look during the week 15:05:26 i've not had time to look either 15:05:47 Also I've spawned 3VMs aio to play with haproxy patches, but haven't started playing much 15:06:37 We've breifly discussed yestarday where it's worth to pay attention - like handler dynamic naming and virtual groups 15:07:46 It's also absolutely worth splitting changes of haproxy_services format and adoption for splitting configs into 2 patches 15:08:58 As of now I tried sandbox without LE, and it seems like approach overall working. The only thing that came to my mind - it will be way trickier to expand pool of haproxies or add some new host 15:09:23 as basically you'll need to run setup-everything. Might be not a big deal given documentation and working tags though 15:16:29 I've also tried bumping mariadb version to 10.11 and it obviously didn't "just worked" as package names seems to have changed. 15:17:05 And now it seems to be aligned with distro names ie not having major version at the end 15:18:39 It seems they've stopped doing that since 10.8. So some work is needed there for sure 15:20:18 I also need these 2 features to inventory land or have some decent alternative to them for 2023.1 15:20:25 #link https://review.opendev.org/c/openstack/openstack-ansible/+/870113 15:20:27 #link https://review.opendev.org/c/openstack/openstack-ansible/+/869762 15:20:55 As we rely on them as of today in concept deployment 15:21:16 (which should become prod by summer) 15:22:53 Another possibly sad topic is uwsgi. I assume some might have read MLs regarding this. In short uwsgi is in maintenance only as of today (and for the last year) 15:28:17 There're discussions ongoing on what should we all do with regards to that and if there should be some tested by default wsgi backend 15:28:35 Or at very least I'd love that this was a thing ^ 15:32:30 o/ 15:32:46 I was thinking that at worst it might be worth renaming uwsgi role to jsut wsgi and have an option to setup gunicorn as well 15:35:12 sorry also in another meeting right now 15:35:22 no worries 15:37:17 I think that's kind of it from my side. I don't think we have any new bugs to discuss 15:37:41 But reviews on stable branches backports are super welcome to get new releases 15:41:09 i will also try to look at the haproxy stuff 15:41:20 i'd like to try the add_host approach 15:45:22 So basically get rid of delegate_to approach? 15:47:08 The thing I'm concerned about, is that with dynamic group, when adding haproxy hosts to glance_api, for example, it will result in placing template on hosts that are not intended for that. But maybe indeed there's a way for that 15:47:42 I was trying to focus on handlers thing now as it's quite annoying as well. 15:48:54 at least concept overall is working from what I can tell. 15:54:20 i was hoping that `add_host` would let you do new_group = haproxy_all hosts [also member of glance_all] 15:54:43 then be able to run against `new_group` which would just be haproxy_all hosts but also with glance vars 16:03:03 well, then role will run against all hosts in this new group, isn't it? 16:03:50 So it will deploy haproxy.conf on both haproxy_all and glance_all 16:05:06 i don't know - i need to make some quick test playbook to see 16:06:35 Maybe, if add haproxy_all to glance_all temporary instead of new_group and run against haproxy_all will do the trick... But it kind of depends in glance_all is group that consist of children or not... 16:06:55 but yeah, worth checking this for sure 16:06:58 #endmeeting