16:01:58 <johnsom> #startmeeting Octavia
16:01:58 <opendevmeet> Meeting started Wed Oct 15 16:01:58 2025 UTC and is due to finish in 60 minutes.  The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:01:58 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:01:58 <opendevmeet> The meeting name has been set to 'octavia'
16:02:16 <gthiemonge> o/
16:02:16 <danfai> o/
16:02:17 <raineszm> o/
16:02:32 <johnsom> #topic Announcements
16:03:41 <johnsom> Reminder the PTG is coming up in a few weeks
16:04:17 <johnsom> #link https://etherpad.opendev.org/p/oct2025-ptg-octavia
16:04:28 <gthiemonge> Yeah!
16:04:29 <johnsom> That is the etherpad. Please add any topics you would like to review.
16:05:07 <johnsom> Also, I approved the patch to move Caracal 2024.1 to unmaintained for Octavia.
16:05:23 <johnsom> #link https://review.opendev.org/c/openstack/releases/+/963581
16:05:55 <johnsom> I was able to get all of the open back ports in (aside from the one that had a missing feature dependency).
16:06:20 <johnsom> Any other announcements this week?
16:06:30 <gthiemonge> nop
16:06:36 <danfai> nop
16:06:47 <johnsom> Great, moving on
16:06:48 <johnsom> #topic Brief progress reports / bugs needing review
16:07:33 <gthiemonge> as I reported earlier today, our pep8 job is broken, due to a bug in recent pylint releases (4.0.x), I proposed a patch to skip these releases
16:07:34 <johnsom> I have been working on the stable branches to get things merged. Someone also proposed a stable branch release that I held as we still had some things to merge, but that should go soon.
16:08:24 <johnsom> I have also been working on the rate limiting feature. Sadly I got pulled away with downstream issues a bit this week, so I have not yet been able to push the API-REF updates. Hopefully this week.
16:08:53 <gthiemonge> I'm also working on adding support for centos 10 stream, I found and fixed a bug in diskimage-builder, my latest smoke test patch passed!
16:08:54 <johnsom> On the pep8 issue, I workflowed the patch that excludes those versions of pylint as it is pretty clear that is a regression
16:09:24 <johnsom> Nice! Once we get CentOS 10 stream going we should update the FIPS jobs to use it.
16:10:18 <johnsom> Any other progress updates?
16:10:45 <danfai> not from me, was busy otherwise, sorry
16:11:14 <johnsom> Of course. We all have many things going on...
16:11:27 <johnsom> #topic Open Discussion
16:11:56 <johnsom> I will be travelling next week and will not be able to host this meeting. Do we want to cancel or does someone else want to run it?
16:12:48 <gthiemonge> I think we can cancel it
16:13:08 <johnsom> Ok, I will send a note out to the mailing list.
16:13:17 <johnsom> I will be back for the PTG. lol
16:14:11 <raineszm> Does anyone hav ea second to discuss https://review.opendev.org/c/openstack/octavia/+/780215?
16:14:22 <raineszm> Sorry, got p;ulled away for a second
16:14:36 <johnsom> Yes, this is open discussion, so now is the time!
16:15:23 <johnsom> Ah, this is a complicated problem....
16:15:26 <raineszm> So I was looking into the reviews on patch 5 and patch 6 and after some digging I don't think it's in general feasible to timeout plugins that get stuck in a loop
16:15:50 <raineszm> Basically in the example given, the ovn-plugin spins up its own thread and there's in general no way for us to force that thread to exit
16:16:11 <raineszm> The best we could do was start the plugin in a new process, but that has its own issues.
16:16:43 <johnsom> Right, and that provider driver has historically had a lot of problems
16:16:46 <raineszm> So it might be best to just try to go with something like ps4 to at least fix the original problem
16:17:08 <raineszm> Ah. Yeah. There's only so much that can be done to protect from a driver plugin misbehaving
16:17:22 <johnsom> There are ways we can force kill a run away driver thread. It's just.... hard
16:18:30 <raineszm> Right. And platform dependent, no?
16:18:53 <gthiemonge> I think that to make the timeout work with ovn-provider, ovs/ovs client needs to be async-compatible. asyncio.wait_for works only with async functions
16:18:58 <johnsom> Yes, though should work on linux just fine
16:19:21 <raineszm> Right. The plugin doesn't hand cancellation
16:19:29 <raineszm> *handle
16:19:30 <johnsom> Yeah, if he is going down the asyncio path....  that is maybe not the best
16:20:02 <raineszm> I looked at using asyncio, multiprocessing, and futures
16:20:23 <raineszm> And yeah, we'd need to track threads started by the plugin and kill them by hand after a timeout.
16:20:48 <gthiemonge> ok then maybe ps4 is fine, and we should ask the providers to not block in case of issues during the init ;-)
16:21:05 <raineszm> That's what I was thinking.
16:21:10 <raineszm> Open to other suggestions though
16:23:14 <gthiemonge> well I guess it will be ps4
16:24:45 <raineszm> That works for me. Should I rebase and repush?
16:25:29 <gthiemonge> yes please
16:25:52 <johnsom> Please feel free to propose it. It would be best if you include an explanation note in the comments as well as some folks involved in that patch are not here today.
16:26:05 <johnsom> Thank you!
16:28:05 <johnsom> Any other topics for open discussion today?
16:28:37 <danfai> not from me
16:29:05 <gthiemonge> no
16:29:30 <johnsom> Ok, thank you all for joining. See you at the PTG!
16:29:42 <johnsom> #endmeeting