16:00:48 <johnsom> #startmeeting Octavia 16:00:48 <opendevmeet> Meeting started Wed Sep 24 16:00:48 2025 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:48 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:48 <opendevmeet> The meeting name has been set to 'octavia' 16:01:07 <gthiemon1e> o/ 16:01:10 <johnsom> #topic Announcements 16:01:17 <johnsom> Hi everyone 16:01:27 <johnsom> Next week is release week for Flamingo! 16:02:04 <johnsom> Also note, the PTG is coming up. The schedule is now open. 16:02:38 <johnsom> I plan to book a couple of hours on Wednesday a bit earlier than this. 16:02:42 <johnsom> Does that work for everyone? 16:03:13 <gthiemonge> yeah, that looks good 16:03:17 <johnsom> I am thinking two hours earlier than this. 16:03:22 <opendevreview> Merged openstack/octavia master: reno: Update master for unmaintained/2023.1 https://review.opendev.org/c/openstack/octavia/+/934790 16:03:28 <johnsom> Ok, sounds good. 16:04:03 <johnsom> #topic Brief progress reports / bugs needing review 16:04:51 <johnsom> I have mostly been stuck doing downstream work, but I have got a few reviews in as well. 16:05:19 <johnsom> I plan to focus on the rate limiting work today (finally) 16:05:33 <gthiemonge> I rebased my work on the tempest tests for SCTP, I'm trying to integrate the tests in the "regular" scenario job, but now the duration is really close to the timeout value 16:06:05 <gthiemonge> maybe a good topic for the PTG: can we remove/merge some tests? 16:06:12 <johnsom> Sigh, ping me when you think you have a setup that is stable. It would be great to get that in 16:06:44 <johnsom> Yeah, I agree, we should review the list of tests that are running. It's been a while 16:07:23 <gthiemonge> for instance we have basic tests and hm tests, I think that the providers should support HMs now so maybe the basic tests are no longer needed 16:08:41 <johnsom> Hmm, yeah, I know OVN didn't support health monitors for a while. That might be an artifact of that 16:10:17 <johnsom> Any other progress updates? I do have a bug I would like to discuss today 16:10:28 <gthiemonge> I think that's all for me 16:10:42 <johnsom> #topic Bug review 16:11:00 <johnsom> #link https://bugs.launchpad.net/bugs/2125573 16:11:43 <johnsom> This is an interesting one. It's a behavior in OVN that is causing problems, but raises a good question. 16:13:06 <johnsom> So in the early days (sounds so old, lol), the allowed address pairs port had to be admin down, otherwise neutron tried to pin it to one compute host. 16:13:44 <johnsom> This caused a lot of problems since it is not a "real" port, but just an IP address alias. 16:14:17 <johnsom> This bug implies this may no longer be the case with OVN. 16:14:27 <johnsom> Anyone have any thoughts on this? 16:15:08 <johnsom> I haven't tried creating an AAP port with admin up in many years, so I have no idea if this has changed 16:15:21 <gthiemonge> not really... the VIP port should be "type: virtual" in OVN, we need to check if the admin_state=False is a condition for neutron to set this type 16:16:16 <johnsom> Ok, I wanted to bring it up here to see if there were any thoughts on it. 16:17:19 <johnsom> I will add some comments and then add neutron to the affected project list. I think this is something to be resolved on the neutron side, but I will keep octavia on the list just in case we do need to change that behavior. 16:17:30 <gthiemonge> +1 16:17:38 <johnsom> #topic Open Discussion 16:17:47 <johnsom> Any other topics today? 16:17:58 <gthiemonge> yes 16:19:04 <gthiemonge> I've seen multiple patches for the infamous one-liner that moves the interfaces from the "root" network namespace to the right namespace 16:19:07 <gthiemonge> like this one 16:19:12 <gthiemonge> https://review.opendev.org/c/openstack/octavia/+/961406/4/octavia/amphorae/backends/agent/api_server/templates/amphora-netns.systemd.j2 16:20:02 <gthiemonge> my concern is that this line is unreadable and unmaintainable, so I would like to suggest to add a script to the amphora image, it could be shell or python, it would make it easier to maintain 16:20:45 <gthiemonge> I'm sweating each time I try to understand it 16:20:56 <gthiemonge> any opinion on this? 16:21:16 <gthiemonge> (I already suggested it in the review but I didn't give a -1) 16:21:39 <johnsom> lol, that has been a bash work of art 16:22:14 <johnsom> I am ok with making it a script. 16:22:32 <johnsom> I think it's only really used in the VM reboot scenario right? 16:22:52 <gthiemonge> another attempt here 16:22:56 <gthiemonge> https://review.opendev.org/c/openstack/octavia/+/949630 16:22:58 <gthiemonge> right 16:23:21 <gthiemonge> when the interfaces are already plugged when the amphora-haproxy namespace is created 16:23:37 <johnsom> Yeah, I have no problem with that. Frankly python might be cleaner / easier to read than doing the fun in bash 16:23:51 <danfai> As I was touching the 949630, I was also wondering instead of that one-liner, if there is maybe a script to be executed instead of 13 lines of systemd ExecStart 16:25:05 <gthiemonge> danfai: +1 16:25:11 <johnsom> So there are two +1's to changing that systemd service 16:26:37 <danfai> Would it be good enough for you, if the one-lienr becomes a bash script, that just makes the one-liner a multi-liner and more readable in a script? 16:27:23 <gthiemonge> I think if it becomes a script, we need to make it readable and maintainable, so basically we need to refactor it 16:27:36 <danfai> What I mean is, create a shell script that gets added to the amphora and this systemd file executes the script instead of the ExecStarts 16:28:10 <danfai> +1 16:28:13 <gthiemonge> I think it can include all the execstarts 16:28:52 <danfai> Is the amphora-interface script used somewhere else? 16:29:28 <gthiemonge> yeah it's called by the amphora-agent when a new interface is plugged 16:29:44 <gthiemonge> https://github.com/openstack/octavia/blob/e414abb9e596241fb110549ace8c472d63a6f34e/octavia/amphorae/backends/agent/api_server/osutils.py#L89 16:31:17 <danfai> ah, ok, I see 16:31:53 <danfai> Not sure I have enough time currently, but if I find a few seconds, I can have a look at writing a bash script with a bit more comments as a start 16:32:11 <gthiemonge> ack 16:32:32 <johnsom> Thank you! 16:33:56 <johnsom> Any other topics today? 16:34:05 <gthiemonge> no, thanks! 16:34:14 <danfai> not from me 16:35:02 <johnsom> Ok, thank you all and have a great week! 16:35:16 <johnsom> #endmeeting