16:01:02 <johnsom> #startmeeting Octavia 16:01:02 <opendevmeet> Meeting started Wed May 21 16:01:02 2025 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:02 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:02 <opendevmeet> The meeting name has been set to 'octavia' 16:01:17 <danfai> o/ 16:01:23 <gthiemonge> o/ 16:01:28 <johnsom> Hi everyone 16:02:33 <johnsom> The only announcement I have is we have passed the Flamingo milestone 1. 16:02:47 <gthiemonge> wow 16:03:26 <johnsom> The next milestone is at the end of June 16:03:34 <johnsom> Any other announcements this week? 16:03:41 <gthiemonge> nop 16:03:45 <danfai> nop 16:04:03 <johnsom> #topic Brief progress reports / bugs needing review 16:04:47 <johnsom> I am currently playing whack-a-mole with setuptools issues, though it seems they aren't impacting Octavia as much as they are other projects. 16:05:00 <johnsom> I fixed a bug with disabling ssh in the amphora image: 16:05:07 <johnsom> #link https://review.opendev.org/c/openstack/octavia/+/949678?usp=search 16:05:40 <gthiemonge> I rebased the (huge) Active-Active BGP patch chain, I'm reviewing the code, I wrote it long time ago, I need to refresh my memory 16:06:09 <johnsom> I also fixed a bug we talked about at the PTG where if the user tries to create a VIP with an IP already in use, the error is horrible. This required an octavia-lib fix too as it was introduced when we added provider drivers. 16:06:15 <gthiemonge> I'll probably create an etherpad where I will throw some ideas about this feature, so we will be able to discuss them 16:06:31 <johnsom> #link https://review.opendev.org/c/openstack/octavia-lib/+/949546?usp=search 16:06:31 <johnsom> #link https://review.opendev.org/c/openstack/octavia/+/949548?usp=search 16:07:33 <gthiemonge> https://review.opendev.org/q/topic:%22bgp-exp%22 16:07:40 <gthiemonge> ^ the active-active patchchain 16:08:01 <johnsom> Finally, I think neutron found a way to continue to test the fips jobs on centos-9-stream, so I'm trying that out. Short story is requirements team dropped support for python3.9 that is what centos-9-stream uses 16:08:18 <danfai> I was mostly busy downstream. There were some issues previously related to quota getting out of sync, so I made a housekeeping job, that can be reviewed. 16:08:23 <johnsom> Nice, I'm looking forward to making progress on that BGP approach 16:08:27 <danfai> #link https://review.opendev.org/c/openstack/octavia/+/949581 16:09:11 <danfai> gthiemonge: happy to contribute to the etherpad. I was checking the recent 4 or so specs a few weeks back 16:09:38 <johnsom> Oh, those quota fix patches never merged.... 16:09:45 <johnsom> #link https://review.opendev.org/c/openstack/octavia/+/886263 16:10:01 <johnsom> #link https://review.opendev.org/c/openstack/octavia/+/886111 16:10:12 <gthiemonge> yeah there are multiple issues with the quotas, a housekeeping task could be a good method 16:10:13 <johnsom> We should prioritize getting those reviewed 16:10:14 <danfai> I'd rather have an manage command to manually fix those as an operator, but house-keeping was a bit easier than writing a new cmd tool 16:11:27 <johnsom> IMO it would be best if we just fixed the bug. 16:11:45 <danfai> +1 , but it is also prone to get out of sync for the next reason 16:11:56 <danfai> * next bug 16:12:21 <gthiemonge> it would be great if your patch could display which quotas were not correct :D 16:12:33 <gthiemonge> and open a launchpad 16:13:00 <johnsom> Yeah, I would hope it would log when it found an issue so we can narrow down where the problem area is 16:13:03 <danfai> It actually does, and fixing them is optional 16:13:28 <gthiemonge> great! 16:13:29 <johnsom> Ok, great. I haven't looked at it yet 16:14:49 <johnsom> Any other updates? 16:14:58 <gthiemonge> nop 16:15:01 <danfai> nop 16:15:12 <johnsom> #topic CI Status 16:16:27 <johnsom> Things are generally ok, but I was looking at the cinder volume backed job today. 16:16:53 <johnsom> It times out a lot. I expect because of the low disk io performance some of the node pool hosts have. 16:17:27 <johnsom> It looks like it is running the full scenario suite. What do you think about narrowing this to a smoke test or two? 16:18:16 <gthiemonge> yeah we don't need all that stuff to test the cinder backend 16:19:17 <johnsom> Ok, I might look into changing the jobs it runs to help it with the timeout issue. We are seeing more people be interested in using this feature, so we should try to work towards making it voting 16:19:25 <gthiemonge> IMHO it would be better if this job uses active-standby + failover test + a few basic tests 16:19:47 <gthiemonge> I can work on it if you want 16:20:46 <johnsom> If you have time that would be great! I'm loaded with setuptools issues (two still open) and a bug in designate at the moment. 16:21:01 <gthiemonge> ack, I'm adding it to my TODO list ;-) 16:21:47 <johnsom> Yeah, I agree, Active/Standby with failover would cover it well. I think the failover does a traffic test, so would have good smoke test coverage 16:22:02 <johnsom> Thanks! 16:22:15 <johnsom> #topic More flexible Anti-Affinity for Octavia 16:22:17 <gthiemonge> cool! 16:22:49 <johnsom> scoopex wanted to talk about this topic. He included a link: 16:22:56 <johnsom> #link https://input.scs.community/JgMcHJ8oQTKk0cC9UzgUjw 16:24:01 <johnsom> This topic has come up before. 16:24:33 <johnsom> In general we have hoped nova would add additional anti-affinity support to cover scenarios beyond just the host. 16:24:54 <gthiemonge> if i understand correctly, we had a similar proposal (https://review.opendev.org/c/openstack/octavia/+/923939) but this new one depends on some new/future changes in nova, right? 16:24:55 <johnsom> For example cross rack, room, site, etc. anti-affinity 16:25:25 <johnsom> We have tried to keep Octavia out of the scheduling space and lean on nova to provide that capability. 16:26:24 <danfai> 923939 was about using nova's AZ to distribute the amphoras iirc 16:26:29 <johnsom> I think so, I'm not very familiar with the 923939 patch 16:26:48 <gthiemonge> right, this new approach is more flexible, not only for AZs 16:27:44 <johnsom> I know there was one group that wanted rack based anti-affinity, so a flexible option would be nice. 16:27:56 <danfai> downstream we patched nova to have similar filters, but I guess without having support in upstream nova for more granular filters, octavia cannot really use more then AZs 16:28:03 <johnsom> That said, I think the main use case I see is cross AZ 16:28:58 <johnsom> Cross AZ also brings in the stretched L2 vs BGP issue (hint, BGP is the right answer IMO) 16:29:44 <johnsom> It seems scoopex isn't here today, so maybe we should take homework to review both of these and be ready to discuss next week 16:30:11 <gthiemonge> yeah, this is a really interesting topic 16:30:23 <johnsom> +1 it's a good one 16:30:48 <danfai> +1 16:31:09 <johnsom> #topic Open Discussion 16:31:14 <johnsom> Any other topics today? 16:31:47 <gthiemonge> not from me 16:31:47 <danfai> not from my side 16:32:25 <johnsom> Ok, thank you all for the discussion. See you next week! 16:32:38 <johnsom> #endmeeting