16:00:03 <gthiemonge> #startmeeting Octavia
16:00:03 <opendevmeet> Meeting started Wed Jul 31 16:00:03 2024 UTC and is due to finish in 60 minutes.  The chair is gthiemonge. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:03 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:03 <opendevmeet> The meeting name has been set to 'octavia'
16:00:05 <gthiemonge> o/
16:00:07 <johnsom> o/
16:00:14 <tweining> o/
16:01:30 <gthiemonge> #topic Announcements
16:01:35 <gthiemonge> * 2024.2 Dalmatian Release Schedule
16:01:42 <gthiemonge> The feature freeze is in 4 weeks
16:01:51 <gthiemonge> https://etherpad.opendev.org/p/octavia-priority-reviews
16:02:17 <gthiemonge> please prioritize spec and feature patches!
16:03:23 <johnsom> +1, we have some SR-IOV patches still open, including some that need to be backported
16:03:48 <tweining> we have lots of patches still open
16:04:24 <tweining> we also need to get better at following up after patches were updated
16:05:17 <johnsom> In other announcements, some stable branch releases have been proposed
16:05:42 <johnsom> Also, the foundation has student groups available for projects if anyone is interested in mentoring some students.
16:06:01 <gthiemonge> right, and antelope will be unmaintained soon
16:06:50 <johnsom> #link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/62AGYQRPDWANS377DLSU7Q7TLNDNXPAK/
16:07:12 <tweining> good that Red Hat will start maintaining right in time then ;)
16:07:30 <tweining> *downstream of course
16:09:26 <gthiemonge> any other announcements?
16:10:15 <tweining> no, only that we didn't meet our goal from last week
16:10:20 <johnsom> Oh, the "E" series schedule has been announced as well
16:10:56 <johnsom> #link https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/V3TU7EZ7QYHI2N57YG35RVUIOAMGDBKY/
16:11:06 <tweining> I mean to integrate at least one spec patch. but there were some reviews at least
16:13:02 <gthiemonge> yeah I have updated the SG on VIPs spec after Michael's feedback, now I'll look at yours
16:14:27 <gthiemonge> #topic CI Status
16:14:33 <gthiemonge> an update here!
16:14:51 <gthiemonge> The fix for the IPv6 tempest job merged today!
16:14:59 <gthiemonge> thank you johnsom!
16:15:06 <johnsom> +1, good to have that running again
16:16:48 <tweining> gthiemonge: https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/921402 (and the follow-up patch) still need approval
16:18:06 <gthiemonge> +1
16:18:08 <gthiemonge> right
16:19:49 <gthiemonge> #topic Brief progress reports / bugs needing review
16:20:25 <johnsom> I have been working on a path to be able to test SR-IOV in the zuul gates.
16:20:27 <tweining> I have nothing else for today.
16:20:58 <johnsom> It's not clear if opendev will be able to have nodepool instances with the right qemu driver in them due to the clouds that provide the resources.
16:21:34 <johnsom> The alternative would be to nest devstack one VM layer deeper, which is not ideal
16:22:05 <tweining> I agree, more overhead
16:22:36 <johnsom> When I set this up local, the igb driver provides 7 VFs by default, which I think will work nicely for the tests.
16:22:55 <johnsom> Most real nics provide up to 255, fyi
16:25:34 <gthiemonge> interesting!
16:30:16 <gthiemonge> #topic Open Discussion
16:30:47 <johnsom> I would like to talk about this patch that came in this morning:
16:30:49 <johnsom> #link https://review.opendev.org/c/openstack/octavia/+/925327
16:31:37 <gthiemonge> yeah I saw it
16:32:06 <johnsom> I have concerns about adding third-party provider driver code to the main repository.
16:32:39 <gthiemonge> it's related to https://review.opendev.org/c/openstack/ovn-octavia-provider/+/925324
16:32:45 <johnsom> Looking at the code, it is accessing Octavia internals and the database
16:32:56 <gthiemonge> I agree
16:33:06 <johnsom> Most of it is Octavia related code, with one line that ties it specifically to OVN
16:33:34 <gthiemonge> I think they're trying to avoid a limitation of the octavia_lib api: retrieving the list of LBs
16:34:15 <tweining> mmh, it's still W-1 so wouldn't even take notice of it normally
16:35:41 <gthiemonge> johnsom: do you want to leave a comment about that?
16:36:25 <johnsom> Well, I was trying to decide what comment to leave....
16:36:45 <johnsom> Option 1, this should go in the OVN provider repo
16:36:57 <johnsom> Option 2, it should be made driver agnostic
16:37:25 <johnsom> Option 3, if you need a list of your own LBs, why can't you query the Octavia API?
16:37:32 <gthiemonge> for 1. I think the ovn provider cannot import Octavia modules directly
16:37:46 * johnsom notes, option 3 is documented in the driver developer guide
16:38:10 <johnsom> This is true, the provider should not for sure
16:38:48 <gthiemonge> yeah mixing Octavia API and octavia_lib is possible
16:40:52 <johnsom> The other issue with that patch is it's creating a new provider driver method that doesn't exist today
16:41:12 <johnsom> I.e. isn't documented, isn't in the base class, etc.
16:42:47 <gthiemonge> so I would be in favor of Option 3. Option 2 is also possible but I'm not sure it would be useful for other provider drivers
16:43:20 <johnsom> One could argue that if one driver needs it, others might too
16:43:43 <gthiemonge> it's a WIP patch, but in order to be proposed they would need to add the method to the driver framework
16:44:12 <johnsom> For sure.
16:44:12 <tweining> if you're concerned, maybe we should comment that this should be discussed further either here or on the mailing list
16:44:24 <johnsom> I just wanted to start a discussion about this
16:44:35 <tweining> I would prefer option 3 as well BTW, because it seems to have the least impact
16:45:32 <gthiemonge> option 3 would be more backportable too
16:45:37 <johnsom> It also means less DB code to maintain
16:47:56 <johnsom> Ok, I think I have enough thoughts and input to comment on the patch
16:48:06 <johnsom> Thank you for the discussion
16:48:09 <gthiemonge> thanks johnsom
16:49:36 <gthiemonge> do you have any other topics?
16:49:46 <johnsom> I don't
16:50:04 <tweining> I already said that I don't either
16:50:11 <gthiemonge> ok!
16:50:16 <gthiemonge> thank you guys!
16:50:24 <gthiemonge> #endmeeting