15:08:08 <noonedeadpunk> #startmeeting openstack_ansible_meeting 15:08:08 <opendevmeet> Meeting started Tue Nov 5 15:08:08 2024 UTC and is due to finish in 60 minutes. The chair is noonedeadpunk. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:08:08 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:08:08 <opendevmeet> The meeting name has been set to 'openstack_ansible_meeting' 15:08:12 <noonedeadpunk> #topic rollcall 15:08:25 <noonedeadpunk> sorry, I got too deep into rabbitmq 4.0 so missed the time 15:08:29 <noonedeadpunk> o/ 15:08:33 <jrosser> o/ hello 15:09:12 <jrosser> feels pretty quiet this week 15:10:09 <noonedeadpunk> # topic office hours 15:10:14 <noonedeadpunk> #topic office hours 15:10:31 <noonedeadpunk> yeah, week seems to be very quiet 15:10:44 <noonedeadpunk> I was working on mariadb and rabbitmq major upgrades 15:10:57 <noonedeadpunk> and these not going very streight so far 15:11:11 <noonedeadpunk> first thing with mariadb - is that we need to switch repos first 15:11:42 <noonedeadpunk> which is a bit more tricky to do in CI 15:13:31 <jrosser> ah yes that will not be so obvious with switching the mirrors 15:14:43 <noonedeadpunk> right now most tricky seems to be on what to do with the upgrade job 15:15:28 <noonedeadpunk> as my plan was to comment out mirror, merge switch, return back new mirror when available 15:15:44 <noonedeadpunk> ie https://review.opendev.org/c/openstack/openstack-ansible/+/934038 15:16:08 <noonedeadpunk> but then it seems we need to disable mirror somewhere during upgrade 15:18:14 <noonedeadpunk> with rabbitmq I'm still playing kinda. as sometimes I do catch upgrade failure, alike to what I saw in CI 15:18:31 <noonedeadpunk> but other times - I'm not. It feels there're just too many dependencies to account for 15:18:47 <noonedeadpunk> and I'm missing some from time to time 15:20:03 <noonedeadpunk> And I also plan to propose patches to cover deb822 format for left roles 15:20:08 <noonedeadpunk> (later this week) 15:20:11 <jrosser> you mean package dependancies of rabbitmq? 15:20:29 <noonedeadpunk> no, patch dependencies 15:20:35 <jrosser> ah ok 15:20:58 <noonedeadpunk> as there's this https://review.opendev.org/c/openstack/openstack-ansible/+/934042 and then that https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/931801 15:21:04 <noonedeadpunk> and then there was smth else to account... 15:21:09 <noonedeadpunk> and that all during upgrade 15:21:52 <noonedeadpunk> but I'd dare to say, that with mariadb and rabbitmq bumps, along with deb822 - we should be able to branch? 15:23:05 <jrosser> that sounds about right yes 15:24:06 <noonedeadpunk> we have 2 broken roles though - adjutant which is broken only with tls - should be doable to fix, and I think it was zun or smth 15:24:32 <noonedeadpunk> nah, it was manila :( 15:24:45 <noonedeadpunk> and it was broken on cephfs part 15:24:51 <noonedeadpunk> *ceph-ansible 15:26:23 <noonedeadpunk> to try and get rid of broken constraints fetch we need to land this and then update role sha in integrated repo: https://review.opendev.org/c/openstack/openstack-ansible-repo_server/+/934046 15:28:45 <noonedeadpunk> fwiw, we have pretty much exactly a month until release deadline 15:29:19 <noonedeadpunk> so I will try to summon folks for reviews once we have most of things ready 15:31:17 <jrosser> i will also try to look over it some more 15:31:27 <jrosser> not managed to spend much quality time on this recently 15:31:33 <noonedeadpunk> we also do have couple of bugs for EL and ceph scenario - https://review.opendev.org/c/openstack/openstack-ansible-ceph_client/+/933606 15:31:56 <noonedeadpunk> with these CI is passing Ceph scenario for https://review.opendev.org/c/openstack/openstack-ansible/+/933592 15:34:20 <noonedeadpunk> but question is - if we actually want to make it voting 15:34:47 <noonedeadpunk> I think I would be fine leaving it as non-voting as well 15:35:16 <jrosser> maybe when it is seen to be stable for some time? 15:35:25 <jrosser> things are difficult enough to merge at the moment 15:35:34 <noonedeadpunk> yeah, fair point 15:36:02 <jrosser> i do think we could improve the u-c trouble for upgrade jobs though 15:36:22 <jrosser> as we discussed before, by making the condition to use the on-disk repo better 15:36:32 <noonedeadpunk> but it kind of collect all dependencies to make the job passing somewhere 15:37:02 <noonedeadpunk> frankly - I guess I'd just extended timeout for now.... as a quick fix 15:37:17 <noonedeadpunk> as I faced same issue in aio yesterday 15:37:26 <noonedeadpunk> so it's not limited to CI 15:37:40 <noonedeadpunk> which is probably a good reality check... 15:38:41 <noonedeadpunk> and probably it's worth backporting even further 15:39:33 <noonedeadpunk> improving usge of on-disk repo still worth looking at ofc 15:42:13 <noonedeadpunk> and other place I've spotted UC fetch issue - was rally role 15:44:10 <noonedeadpunk> but uri, unlike to get_url has timeout of 30 seconds 15:44:17 <noonedeadpunk> (by default) 15:45:17 <opendevreview> Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_rally master: Add retries for UC fetching over HTTP https://review.opendev.org/c/openstack/openstack-ansible-os_rally/+/934144 16:03:02 <noonedeadpunk> #endmeeting