16:00:40 <gthiemonge> #startmeeting Octavia
16:00:40 <opendevmeet> Meeting started Wed Jun 30 16:00:40 2021 UTC and is due to finish in 60 minutes.  The chair is gthiemonge. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:40 <opendevmeet> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:40 <opendevmeet> The meeting name has been set to 'octavia'
16:00:46 <gthiemonge> Hi
16:01:22 <johnsom> o/
16:01:25 <haleyb> hi
16:02:19 <gthiemonge> #topic Announcements
16:02:25 <gthiemonge> CI is RED today
16:02:51 <gthiemonge> 1) PEP8 job is broken
16:03:07 <gthiemonge> A new pylint release introduced new checkers
16:03:21 <gthiemonge> and some older checkers detects new errors
16:03:34 <gthiemonge> I proposed a fix, I think it's already CR+2 W+1
16:03:45 <johnsom> Yay for linters....
16:03:47 <gthiemonge> #link https://review.opendev.org/c/openstack/octavia/+/798811
16:04:17 <gthiemonge> I got it when rechecking a patch after the neutron SG rule deletion issue was merged :/
16:04:45 <gthiemonge> 2) octavia-tox-functional-py37-tips is unstable
16:04:57 <gthiemonge> there's a timeout issue when testing the DriverAgent
16:05:09 <gthiemonge> Probably related to the slowness of the CI host
16:05:18 <gthiemonge> I have a workaround, I'll propose it after the meeting
16:05:33 <johnsom> Do you have a link to the logs of one of these failures?
16:05:36 <gthiemonge> but I still don't understand why it occurs only on the -tips job
16:06:06 <gthiemonge> #link https://zuul.openstack.org/build/d7abd5b0ef7b49c69e3325313fb46603
16:06:22 <johnsom> This is a "learning opportunity" as that is my code. grin
16:06:24 <johnsom> Thanks
16:06:32 <gthiemonge> basically there's a settimeout(5) on a socket in the driver_lib
16:06:42 <gthiemonge> and we're hitting the timeout
16:07:13 <gthiemonge> in my env, receiving a message in the functional tests takes less than 200ms
16:07:21 <rm_work> there's some stuff that periodically I realize is not mocked that probably should be in certain situations
16:07:26 <rm_work> this is functional tho?
16:07:29 <gthiemonge> I managed to reproduce it with the cpulimit tool
16:07:40 <gthiemonge> functional yes
16:07:44 <rm_work> hmm k
16:08:20 <johnsom> Yeah, it's a very simple unix domain socket protocol, so ... interesting that it can't send data in 5 seconds
16:08:51 <gthiemonge> yeah i agree
16:10:40 <gthiemonge> any other announcements?
16:10:57 <johnsom> PTG is scheduled and registration is open.
16:11:24 <gthiemonge> Oh, I missed that :D
16:11:28 <johnsom> #link https://www.openstack.org/ptg/
16:12:08 <gthiemonge> johnsom: thanks
16:12:24 <rm_work> I need these to be in-person again T_T
16:12:47 <gthiemonge> in... person... what is that?
16:13:06 <johnsom> We are all virtual AIs
16:14:06 <johnsom> There is also a lively discussion of community goals for the next release on the discuss list.
16:14:30 <johnsom> #link http://lists.openstack.org/pipermail/openstack-discuss/2021-June/023350.html
16:14:49 <johnsom> If you are interested in the goal process
16:17:58 <gthiemonge> #topic Brief progress reports / bugs needing review
16:19:08 <gthiemonge> I found an issue regarding haproxy worker cleanup in active standby
16:19:19 <gthiemonge> I opened a story that describes the behavior:
16:19:27 <gthiemonge> #link https://storyboard.openstack.org/#!/story/2009005
16:19:34 <gthiemonge> and I proposed a patch:
16:19:44 <gthiemonge> #link https://review.opendev.org/c/openstack/octavia/+/797882
16:20:17 <johnsom> That one is super interesting and a good find!
16:20:24 <gthiemonge> I wanted to get some reviews, but it seems it's already W+1
16:20:33 <johnsom> In theory broken for quite a while.
16:21:12 <gthiemonge> yeah the stick tables are not synchronized after a config reload
16:21:19 <johnsom> I still need to go back and read up as I thought that was done over unix domain sockets, so wouldn't need lo, but that is probably something I remember wrong.
16:21:56 <johnsom> This goes back to design sessions we had in Seattle with a HAProxy core
16:22:10 <gthiemonge> I read somewhere that haproxy uses the peers in the peer list even for the local sync
16:22:47 <johnsom> I have been working on changing how we use tempest service clients to align with some "new-ish" guidelines for tempest plugins.
16:22:59 <johnsom> #link https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/797715
16:23:21 <johnsom> Huge thanks to g-mann (no need to ping him) for helping with that.
16:23:22 <gthiemonge> johnsom: and it started with a one-line patch
16:23:28 <johnsom> It did.... sigh
16:24:05 <johnsom> A conversation in #openstack-qa with the ironic team about service client credentials....
16:24:48 <johnsom> Ah, well, that should be pretty close if not good. I need to re-review today
16:25:22 <johnsom> Otherwise my time has been on reviews and Designate things
16:25:28 <gthiemonge> I'm still working on the two-node job, a commit was in review, but it was broken by the switch to ML2/OVN
16:25:55 <gthiemonge> with some help from the neutron folkks, I managed to fix the last issue (floating ips not reaachable)
16:26:04 <gthiemonge> #link https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/773888
16:26:29 <gthiemonge> but now: TIMED_OUT
16:26:30 <gthiemonge> great
16:27:03 <johnsom> Grin
16:27:41 <johnsom> We should look into the "parallel" devstack install code. I thought that was default now, but maybe not? It would in theory stop devstack from installing each node sequentially.
16:29:03 <johnsom> Hnmm, looks like that is working there, so... maybe another issue
16:29:19 <gthiemonge> yeah
16:29:39 <gthiemonge> it looks like controller2 is ready 50 min after the start of the job
16:29:45 <opendevreview> Merged openstack/octavia stable/wallaby: Fix race conditions between API and worker DB calls  https://review.opendev.org/c/openstack/octavia/+/798255
16:29:50 <johnsom> Hmm yeah, so maybe not
16:30:40 <gthiemonge> I'll take a look at it
16:30:42 <johnsom> 50 minutes.... I wonder if devstack still says "will be done in 10 minutes". lol
16:32:01 <gthiemonge> #topic Open Discussion
16:32:10 <gthiemonge> any other topics?
16:34:24 * haleyb is still neck deep in the ovn provider, so nothing else from me
16:35:08 <gthiemonge> ok folks
16:35:11 <gthiemonge> thank you!
16:35:20 <gthiemonge> #endmeeting