16:01:05 #startmeeting Octavia 16:01:05 Meeting started Wed Dec 9 16:01:05 2020 UTC and is due to finish in 60 minutes. The chair is rm_work. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:06 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:08 The meeting name has been set to 'octavia' 16:01:13 Opps, yes, thanks 16:01:21 Hi 16:01:23 o/ 16:01:31 Hello 16:01:41 #topic Announcements 16:01:56 Hi guys! I just want to introduce myself. My name is Martin Chlumsky and I work at Ubisoft. We have several hundred loadbalancers in Octavia (Stein) in several Openstack deployments. I've been tasked to improve our Octavia deployments (upgrade to Ussuri, better monitoring, etc...) so you'll see me in here asking questions and I'd like to contribute 16:01:56 where I can (time permitting). :) 16:02:02 After today I am on vacation until January 4th 16:02:26 #chair johnsom 16:02:26 Current chairs: johnsom rm_work 16:02:27 mchlumsky Welcome! 16:02:45 #topic Announcements 16:02:53 Opps, lol 16:03:01 now you're good :D 16:03:06 I will repeat myself: After today I am on vacation until January 4th 16:03:23 johnsom: enjoy! 16:03:25 mchlumsky This channel is a great resource to ask questions about Octavia. 16:03:31 Similar for me after next week -- I think i have one more meeting in me 16:04:08 johnsom I noticed already giot a few answers. thanks! :) 16:04:14 Nice, enjoy your break 16:04:57 Any other announcements this week? 16:05:48 #topic Brief progress reports / bugs needing review 16:06:15 I have mostly been working on reviews and poking at the bandit issues on stable/stein. Otherwise downstream work 16:07:24 Can we start using the priority review etherpad? I know that some people have added patches to the awaiting prioritization list... 16:07:53 yeah I really need to get back in the groove of reviews 16:07:55 Yeah, good question. Milestone 1 has now passed 16:08:09 my guess is the first opportunity i'll have to do so will really be in January 16:09:13 gthiemonge I would say feel free to freshen up the priority review etherpad and get it started for Wallaby 16:09:55 ok, I'll update the list tomorrow 16:10:52 * johnsom likes not being PTL this cycle and still delegating. grin 16:11:12 Any other updates this week? 16:11:32 Thank you all for helping review the stable branch patches. I think we have made good progress this week 16:11:51 #topic Stable/Stein gate jobs are broken due to bandit py27->py3x transition 16:12:21 As you may have seen on the openstack-discuss e-mail list, bandit released a version that no longer supports python2.7. 16:12:39 This broke a number of projects including the stable/stein branch test jobs. 16:13:10 rm_work and I have poked at the issue and I will probably spend a bit of time on that today again. 16:13:26 This is blocking some of our tempest patches from passing. 16:14:00 We will pin bandit on the stein branch to a version that supports python2.7 16:14:19 #topic Open Discussion 16:14:24 Any other topics this week? 16:14:31 I have a story I would like to work on (https://storyboard.openstack.org/#!/story/2008060) as we will likely need this feature. I think it should be labelled as a RFE (but I'm not too sure of the complexity of the change yet so maybe it's a spec?). I might require some guidance but I'll reach out unless you have some vital pieve of information to 16:14:31 give me right now. ;) 16:15:56 Ah, yeah. So, basically you will want to enable changing that setting and then trigger a load balancer failover process to enact the change. 16:17:07 I would advise against trying to hot-change CPU or RAM via nova as it unfortunately triggers an instance reboot, which can cause the security content (TLS keys) to be lost, which will trigger a failover as well. 16:17:30 Would this be done via 2 separate API operations? or can this be done behind the scenes via a PUT to the LB? 16:17:32 Does that align to your thoughts? 16:17:53 You can do it all behind one API call 16:19:22 I won't touch anything on the nova side. I was totally thinking of doing a failover to get this done. I'm just not sure what will happen when you change the topology from active-standby to standalone for example 16:20:07 you'd need to delete the backup basically, and send a new config to the remaining amp, generated with STANDALONE as the type 16:20:11 Well, with the changes I made during Victoria to the failover flow, it should handle the case ok. It will just think that nova lost one of the amphora instances. 16:20:24 but yeah, the failover flow itself should basically just "do that" 16:20:52 new failover flow is great! :thumbsup: 16:21:09 I think I would do a full LB failover. I wouldn't mess with converting an existing instance. Converting would mean a lot of changes inside the amp (keepalived, etc.) that just isn't worth it 16:21:36 ah yeah that's probably cleanest 16:21:47 ok, great. 16:22:13 actually yeah, would it literally just be changing the topo to standalone, and triggering the failover? XD 16:22:34 Yeah, pretty much. The tricky part is handling the rollback correctly maybe. 16:22:35 i suppose that might just work 16:23:01 thanks for making this all so easy for me :p 16:23:20 So, I think this is pretty straight forward and we had a good discussion on it. I'm not sure it needs a full spec to hammer out the details. I would just tag this story as RFE and propose the change 16:24:06 awesome. 16:24:55 Any other topics this week? 16:25:01 yes! here 16:25:17 one of my colleagues is working on a tempest test for TLS 16:25:21 https://review.opendev.org/c/openstack/octavia-tempest-plugin/+/763169 16:25:29 it looks like it's failing on stein 16:26:02 I'm wondering if it's worth it to spend some time to fix stein, or we could just skip the test for some old releases 16:26:37 Well, if the feature was present in stein, we should not skip the test. 16:27:06 how long until we can EOL stein again? :D 16:27:27 we should really be doing one EOL per cycle... 16:27:30 what's up next? 16:27:38 For one thing, I would strongly advise to not change the existing tests, but add additional. 16:27:59 ok, we can check the test in stein... perhaps one backport is missing there 16:28:06 It appears the existing tests were incorrectly modified 16:29:29 I haven't reviewed the patch yet... but yeah the issue might be in the test itself 16:29:55 I can review later, but that is my first comment. Don't change existing tests unless absolutely necessary. 16:30:27 ok thanks 16:30:43 Especially with the TLS tests, it's easy and quick to just create another listener for your new test case. 16:31:03 The current TLS tests are very detailed and still valid IMO 16:32:48 Any other topics to raise? 16:33:59 Ok, thanks everyone. Have a great week and holiday! 16:34:12 johnsom: thank you 16:34:15 #endmeeting