13:00:15 #startmeeting kolla 13:00:15 Meeting started Wed Sep 4 13:00:15 2024 UTC and is due to finish in 60 minutes. The chair is bbezak. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:15 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:16 The meeting name has been set to 'kolla' 13:00:52 #topic rollcall 13:01:01 o/ 13:01:02 o/ 13:01:03 \o 13:01:06 o/ 13:01:08 o/ 13:01:42 \o/ 13:01:48 \o 13:02:02 #topic agenda 13:02:06 * Roll-call 13:02:06 * Agenda 13:02:06 * Announcements 13:02:06 * Review action items from the last meeting 13:02:06 * CI status 13:02:08 * Release tasks 13:02:08 * Regular stable releases (first meeting in a month) 13:02:10 * Current cycle planning 13:02:10 * Additional agenda (from whiteboard) 13:02:12 * Open discussion 13:02:27 #topic CI status 13:02:54 Looks green 13:03:11 Yoga broken 13:03:16 which is suspicious 13:03:19 https://review.opendev.org/c/openstack/kolla-ansible/+/927836 13:03:22 right c8 13:03:23 quick fix 13:03:28 https://review.opendev.org/q/Ib82833168efc2a6e2b4f1178258840d08deb78ef 13:03:31 long fix 13:03:39 yoga is somehow unmaintained :) 13:03:45 Yep 13:03:54 yep, but should be fixed if possible 13:04:01 fixes provided 13:04:04 it will eventually 13:04:08 thx but from unmaintained team, not from kolla team 13:04:42 but it still CI in Kolla project) 13:05:06 IMO not 13:05:41 lets drop unmaintained branches... and the problem is gone 13:05:59 those are a bit limbo state indeed 13:06:02 is there an unmaintained ci irc channel? not that I want to join that :D 13:06:18 SvenKieske: yes #openstack-unmaintained 13:06:31 thanks 13:06:32 ah nice to know, so we can redirect people there I guess? :) 13:06:37 but it's unmaintained and unoffical ;) 13:06:43 :D 13:06:51 it would be /dev/null I guess 13:06:52 the same peoples there) 13:07:19 anyway please merge: https://review.opendev.org/c/openstack/kolla-ansible/+/927836 13:07:22 quick fix 13:07:22 personally, I have no issue with unmaintained branches, I don't need to do anything there, sometimes I get some mail because of a backport, but I can quickly filter those out. 13:07:55 #topic Release tasks 13:08:04 and I'll push folks from unmaintained irc channel 13:08:19 we're R-4 13:08:54 nothing major from the release tasks there 13:09:09 #topic Regular stable releases (first meeting in a month) 13:09:29 I'll handle releases after the meeting 13:09:37 #topic Current cycle planning 13:10:13 I have a - maybe not that encouraging - update with regards to the mirror stuff 13:10:27 Can we discuss https://review.opendev.org/c/openstack/kolla/+/926744 vs https://review.opendev.org/c/openstack/kolla-ansible/+/926632 ...same problem different solution 13:10:27 I've managed to touch a bit OVN BGP agent changes - will need to spawn proper test env for it now. As I've fixed it to be at least deployable 13:10:44 but it is not easily configurable at this stage 13:10:49 but I have ideas 13:11:30 nice :) I followed some of those patches along 13:12:28 kevko: let's wait for mr PTL to be back to discuss those two variants, as one of it is his mnasiadka 13:12:42 the tl;dr for moving rmq or possible other packages into openinfra based mirrors is: I asked over at #opendev and was basically told "it's a bad idea to build your own images, why don't you use upstream" (this is very oversimplified of course, no direct quote) 13:13:19 bbezak: okay, no problem ..but any opinions ? 13:13:27 so they rather would see less mirrored "external" packages imho then adding more. I can understand some of the motivation, but I'm not sure I agree with everything. 13:14:04 not at this point kevko. didn't have time to look there. will do later on 13:14:14 Okay 👍 13:14:43 I did look at both, but I really haven't got a very distinct opinion which I find better, need to think about that some more.. 13:14:51 on their own they both look reasonable 13:15:49 ok, maybe post some of the details if you like to the etherpad SvenKieske 13:16:10 bbezak: okay, will update that (hopefully today) 13:16:37 there's a related point on the additional agenda, but lets discuss this then I guess. 13:16:42 so later I mean 13:16:57 SvenKieske are you suggesting that's a dead end on the mirrors then? Or just that the opendev folks will need more convincing 13:17:55 not sure :D they seemed not really convinced, yes. but my request was not very concrete, maybe if I just propose rabbitmq mirrorring it's not so scary. 13:18:07 I would try one time more I guess and then give up. 13:18:57 I think we could go now to 13:19:00 #topic Additional agenda (from whiteboard) 13:19:19 Merged openstack/kolla-ansible master: CI: also redirect stderr during cert generation https://review.opendev.org/c/openstack/kolla-ansible/+/927530 13:19:27 Rabbitmq magic 13:20:13 yeah, I tried to write up and catch all the patchsets floating around, first off, to have awareness, because I accidently wrote a patchset when there already existed some. 13:20:59 I'm going to add some comments into the whiteboard, basically saying I would like everything to be configurable ;) 13:21:21 please do :) 13:22:10 good SvenKieske, thx. I like us to have good working defaults 13:22:26 with possible other options as mattcrees says 13:22:57 let's discuss that in etherpad as this will be longer discussion I think 13:22:58 For sure, when I say "everything configurable" I don't mean it all stays disabled by default. Just want operators to have the option in case they hit issues CI doesn't catch 13:23:08 yeah sure 13:23:33 I also very munch like the queue_manager thing from kevko, but he was still testing that, does that work now kevko? 13:24:16 No 13:24:21 There is a bug 13:24:23 :/ 13:24:27 It works half 13:24:39 A bug with oslo.messaging, or the KA patches? 13:25:07 Because of containerized environment there is a bad design how to check if the process is same or not 13:25:25 So it's working ..but queues are not reused 13:25:35 yeah, oslo.messaging built it, that this information is shared via /dev/shm afiak 13:25:45 That's not problem I would say 13:25:52 so we would need to expose the same /dev/shm per host for all containers belonging to the same service 13:26:19 yes not really a problem, just annoying I guess :) 13:26:29 Well I like that feature so I will play around this 13:27:03 ok I guess we have consensus to make this stuff configurable by the user with some sane defaults 13:27:07 Could be improved in a follow-up patch I suppose? If it still works, just inefficiently 13:27:45 it might get buggy if the queue names differ, I'm not 100% sure. 13:28:02 Well, now it's not in Kolla ...and I prefer to not give a user an option if it is not working ok 13:28:45 Queue manager is false by default and Kolla not setting to yru for now ...let's leave it as it is now 13:29:04 *true ..sorry mobile 13:29:08 in the meantime I merged your backports from whiteboard SvenKieske, and your quick fix in unmaintained/yoga mmalchuk 13:29:23 thanks 13:29:37 thanks! 13:30:27 let's go swiftly to 13:30:29 #topic Open discussion 13:30:49 Hello. I have noticed that podman part of kolla_container doesn't set KOLLA_SERVICE_NAME environment variable on container creation as docker part does. 13:30:54 Do you think it deservers a fix? 13:31:02 please review backports: https://review.opendev.org/q/Ib9afd84cac1fcbd96f98b4720ea9c6503bbdb124 13:31:22 mhiner: sounds like a simple fix I guess :) 13:31:55 mattcrees: did you happen to get to the root cause why our rabbitmq mirrors still work after https://www.rabbitmq.com/blog/2024/08/11/package-repository-updates ? 13:32:00 yeah, will take care of it soon 13:32:06 will check mmalchuk 13:32:27 thanks 13:32:31 will also look into those mmalchuk 13:32:31 mhiner: I think it should be fixed 13:32:41 SvenKieske no, I haven't found time to look into that yet. 13:33:09 We should probably move off them anyway 13:33:28 Hi, me again with my old patch ... https://review.opendev.org/c/openstack/kolla-ansible/+/920377 :) I really not sure that we have to add something to the documentation at the moment .. 13:33:30 yeah, the blog says: The previously used domain names won't go away, however, this migration would make your setup more future-proof for large infrastructure changes our team expects to happen in the rest of 2024. 13:33:49 so I guess they will go away sooner or later 13:34:37 Ah ok, I'd read that as they were hanging around until Aug 18th. Guess they mean they'll still be available for longer 13:35:53 I'll propose a patch moving to the new ones 13:37:50 I'll check your change chembervint. IIRC there was some queries of healtcheck failing to be added to the doc 13:37:58 and also I want to discuss if do we need a feature - it seams we have need ability to custom systemd unit files for different kolla services. use-cases: 1. config different stuff such as OOM killer strategy and so on. 2. we have to configure dependencies between services - example - restart of ovs_db container should trigger also restart of neutron-openvswitch-agent to sync openflow rules. and actually we have a bunch of useca 13:37:59 ses, where such dependencies could increase realiability of the system. what do you think? 13:38:06 right the mnasiadka comment 13:39:59 Could you please raise a bug in launchpad with description of the feature request chembervint ? 13:40:34 yes, I personally think it would be good to be able to configure systemd unit dependencies properly - if not done already - to address these usecases 13:40:49 yes, of course. Just wanted to ask your first opinion :)) 13:40:57 and yes we should probably have a proper bug for that, very nice :) 13:41:02 we will review it there. as recently we are doing more of bug triaging 13:41:18 ok, we will do bug+commit this-next week 13:42:19 thx 13:43:12 anything more to discuss today? 13:43:50 I guess not. thank you all! 13:43:51 #endmeeting