16:00:13 <gthiemonge> #startmeeting Octavia
16:00:13 <opendevmeet> Meeting started Wed Sep  6 16:00:13 2023 UTC and is due to finish in 60 minutes.  The chair is gthiemonge. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:13 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:13 <opendevmeet> The meeting name has been set to 'octavia'
16:00:16 <gthiemonge> greetings!
16:00:18 <johnsom> o/
16:00:25 <oschwart> o/
16:01:31 <gthiemonge> #topic Announcements
16:01:35 <gthiemonge> * 2023.2 Bobcat Release Schedule: R-4
16:01:42 <gthiemonge> Next week is RC1
16:02:09 <gthiemonge> we expect that it will be our final release for Bobcat
16:02:16 <gthiemonge> If you have important bugfixes to include in the release, speak now
16:03:34 <gthiemonge> or later
16:03:54 <gthiemonge> any other announcements? oschwart? johnsom?
16:04:24 <johnsom> Nope
16:04:36 <oschwart> nope
16:04:38 <gthiemonge> #topic CI Status
16:05:14 <gthiemonge> we got many issues at the begining of the week with grenade (focal was removed from devstack, we were still on it) and zuul (zuul was not triggered when it detected deprecations in the config - like old style regexp that all the projects use)
16:05:24 <gthiemonge> everything is fine now, thanks to the #opendev folks
16:05:35 <gthiemonge> that said, we can fix the deprecations in octavia and octavia-tempest-plugin
16:05:40 <gthiemonge> https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/893705
16:05:42 <gthiemonge> https://review.opendev.org/c/openstack/octavia/+/893704
16:06:22 <gthiemonge> if you want to see what deprecations look like: https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/893705/2#message-585d6ad2aea441b683ba4c2c7afb991325777608
16:08:19 <gthiemonge> and that's it for CI status
16:08:26 <gthiemonge> #topic Brief progress reports / bugs needing review
16:08:45 <gthiemonge> FYI I reported 2 new interesting bugs:
16:08:52 <gthiemonge> failover of ACTIVE_STANDBY LBs can take a lot of time in amphorav1 (when both amps are down)
16:08:56 <gthiemonge> https://bugs.launchpad.net/octavia/+bug/2033894
16:09:02 <gthiemonge> failover of an ACTIVE_STANDBY LB recreate only one amphora when both amps are failing (only >=zed)
16:09:07 <gthiemonge> https://bugs.launchpad.net/octavia/+bug/2033734
16:09:14 <gthiemonge> I'm working on that
16:09:21 <gthiemonge> and I proposed 3 patches:
16:09:28 <gthiemonge> https://review.opendev.org/c/openstack/octavia/+/893537
16:09:37 <gthiemonge> ^ the timeout_dict param was not passed to the correct function, this small patch reduces the duration of the failover
16:09:46 <gthiemonge> https://review.opendev.org/c/openstack/octavia/+/893536
16:09:59 <gthiemonge> ^ skip some amphora API calls if we detect that an amphora is not reachable (also reduce the duration of the failover)
16:10:41 <gthiemonge> (I W-1 this patch becasue I think we can still improve it, it still takes 10 min to recreate an amphora with some specific conditions)
16:10:47 <gthiemonge> https://review.opendev.org/c/openstack/octavia/+/893612
16:10:58 <gthiemonge> ^ don't mark "other" amphorae in ERROR during the failover of an amphora
16:11:16 <gthiemonge> Note1: while one of the bugs impacts only amphorav1, the fixes also improve failovers in amphorav2
16:11:38 <gthiemonge> Note2: the backports of the patches that update amphorav2 will also have to include the changes for amphorav1
16:11:47 <gthiemonge> Note3: I don't think we will merge them before RC1
16:12:07 <gthiemonge> I'm also working on a tempest test that will check this scenario
16:12:53 <oschwart> gthiemonge: lots of fixes, thanks for working on them
16:13:17 <oschwart> I would like to review them when they are ready
16:13:41 <gthiemonge> the first one is ready, it's a trivial patch
16:14:40 <oschwart> I see that it failed on grenade and octavia-v2-dsvm-scenario-non-traffic-ops-jobboard
16:14:48 <oschwart> probably recheck would make it pass now
16:16:27 <gthiemonge> I need to check octavia-v2-dsvm-scenario-non-traffic-ops-jobboard
16:17:41 <gthiemonge> alright
16:17:45 <gthiemonge> #topic Open Discussion
16:17:49 <gthiemonge> I have one topic here
16:18:08 <gthiemonge> I'm working on "Add tests for additional VIPs" (a really old unmerged change)
16:18:13 <gthiemonge> https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/664462
16:18:33 <gthiemonge> the commit adds tests for multi VIP to a new directory: octavia_tempest_plugin/tests/multivip_scenario/
16:18:47 <gthiemonge> does it make sense to have a specific directory/job for this feature?
16:19:00 <gthiemonge> to me that sounds like normal feature, they should be in the scenario dir, what do you think?
16:19:27 <johnsom> Only if we need to run them independently of the other tests for some reason. Like a special setup that would conflict with other tests.
16:20:18 <gthiemonge> I think there's a special setup because we create extra subnets for those tests
16:20:30 <gthiemonge> but this is something that we could always do, for all the tests
16:20:38 <johnsom> Yeah, that shouldn't conflict
16:20:46 <johnsom> Ugh, that patch still has six in it....
16:20:48 <johnsom> old
16:20:55 <gthiemonge> they could be in a dedicated class, with it's own setup method
16:20:58 <gthiemonge> its
16:21:06 <gthiemonge> ootch
16:21:55 <johnsom> Yeah, I would consolidate it into one of the existing test runs. traffic ops maybe?
16:22:49 <gthiemonge> yes
16:23:23 <oschwart> traffic ops sounds good to me for multiple vips
16:23:34 <oschwart> both ipv4 and 6 I guess
16:26:14 <gthiemonge> ack
16:26:29 <gthiemonge> ok, thanks for your feedback guys
16:27:48 <gthiemonge> anything else for this meeting?
16:28:06 <johnsom> I don't have anything
16:28:12 <oschwart> oh yes
16:28:16 <oschwart> let me find the link
16:28:38 <oschwart> I wanted to consult with you guys how to continue this one:
16:28:39 <oschwart> https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/893066
16:29:18 <gthiemonge> I haven't reviewed it yet
16:29:20 <oschwart> it seems like the cert manager tries to parse/retrieve the certificates like if it weren't using noop drivers
16:29:36 <gthiemonge> we have never tried to use certificates in noop tests
16:29:38 <johnsom> That could be
16:29:49 <gthiemonge> so maybe something is missing in Octavia to handle such cases
16:29:56 <johnsom> +1
16:30:01 <oschwart> that's what I believe
16:30:18 <johnsom> This was all an oversight, so I am not surprised we are finding other bits missing/broken
16:30:47 <oschwart> I would like to try working on it, but I will probably need your guidance
16:31:09 <johnsom> Sure, thanks. Feel free to ping me
16:31:19 <gthiemonge> oschwart: I think you need to identify the places where the code fails
16:32:00 <oschwart> johnsom: thanks, and gthiemonge: I think that I found it already, but I need to test it to make sure
16:32:38 <gthiemonge> so we have certificates.cert_manager      = local_cert_manager
16:32:38 <oschwart> regarding milestones - HSTS tests are rebased on top of this patch
16:32:45 <gthiemonge> but maybe we need a noop_cert_manager
16:32:48 <gthiemonge> a new driver
16:33:00 <johnsom> +1, I wondered the same
16:33:36 <gthiemonge> oschwart: there are a few drivers for certificates: https://opendev.org/openstack/octavia/src/branch/master/octavia/certificates/manager
16:33:51 <oschwart> gthiemonge: I might be able to do it, comparing other noop features and certificates drivers
16:33:53 <gthiemonge> the interface is https://opendev.org/openstack/octavia/src/branch/master/octavia/certificates/manager/cert_mgr.py
16:34:08 <gthiemonge> you need to implement this interface in a noop driver that just does some dummy things
16:34:48 <oschwart> gthiemonge: sounds good to me, I should probably open a launchpad for it, right?
16:35:00 <gthiemonge> good idea
16:35:23 <oschwart> ack
16:35:38 <gthiemonge> noop drivers should be really straightforward
16:36:02 <oschwart> I hope so :)
16:36:11 <oschwart> I don't have any more topics beside that one
16:36:25 <gthiemonge> ack
16:36:30 <gthiemonge> anyways, ping us if you need
16:36:41 <oschwart> thanks!
16:38:35 <gthiemonge> ok folks, good discussions!
16:38:41 <gthiemonge> thank you! have a good week!
16:38:48 <gthiemonge> #endmeeting