20:00:01 <johnsom> #startmeeting Octavia 20:00:02 <openstack> Meeting started Wed Oct 10 20:00:01 2018 UTC and is due to finish in 60 minutes. The chair is johnsom. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:03 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:00:05 <openstack> The meeting name has been set to 'octavia' 20:00:22 <johnsom> Hi folks, time for another installment of Octavia this week.... 20:00:34 <nmagnezi> o/ 20:00:51 <johnsom> Quiet around here.... 20:01:16 <nmagnezi> I know cgoncalves was traveling so he might not be around, I don't know for sure 20:01:17 <cgoncalves> hi 20:01:24 <johnsom> I think some folks are off on vacation, so might just be a few of us 20:01:25 <nmagnezi> Mmm never mind :) 20:01:45 <johnsom> #topic Announcements 20:02:03 <johnsom> I don't really have any announcements today. Anyone else have anything? 20:02:18 <cgoncalves_> we released Rocky 3.0.1 20:02:26 <rm_work> o/ 20:02:27 <johnsom> Ah, yes, good stuff 20:03:04 <johnsom> Ok, moving on then 20:03:06 <johnsom> #topic Brief progress reports / bugs needing review 20:03:12 <rm_work> ah, i'll be out for the next 3 weeks 20:03:28 <rm_work> so... see you in a while :P 20:03:46 <johnsom> Ok, good to know. Hope you didn't have plans to visit the beaches of Florida.... 20:04:01 <nmagnezi> lol 20:04:14 <johnsom> I have been busy working on the IPv6 bug. 20:04:19 <cgoncalves_> it's a bit windy there, I heard 20:04:50 <johnsom> This ballooned into a bigger task than planned, but does clean up some long known tech debt in the active/standby implementation. 20:05:11 <johnsom> I have one more issue to address and some more testing before it is ready, but should be done this week. 20:05:33 <nmagnezi> johnsom, yes. I don't think it even worked for me for IPv4, but never got the chance to debug since we had other fires to put out 20:05:41 <johnsom> I have been working in a way that should allow backporting. I really wanted to extend the DB schema, but figured out a way to do it without. 20:06:04 <nmagnezi> johnsom, that's REALLY great 20:06:53 <johnsom> Well, active/standby worked just fine with IPv4 addresses (we are using it widely here), but the issue came in with IPv6. It was caught with our shiny new tempest tests. 20:07:44 <johnsom> After this patch, the active/standby amps will only have a VIP IP if it is the current master. Prior there were extra interface configurations that were not really correct, but didn't hurt things either. 20:07:44 <cgoncalves_> johnsom: does it work well/better now with mixed v4 and v6? 20:07:59 <nmagnezi> johnsom, what I tried to do with this patch is to have only the active node with the IPv4 VIP. But yeah it does work even with the current state 20:08:27 <johnsom> Yes, mixed should still work (need to test that still), but I am not going after getting UDP mixed to work. That will still be future work 20:08:50 <cgoncalves_> cool, that's fine 20:09:49 <johnsom> I might spend a few hours and see if I can get a mixed tempest gate going. The only part I'm not sure about is how to make tempest able to reach an IPv6 address in the gates.... Should be possible, just need to look into it. 20:10:13 <johnsom> We really need an ipv6 traffic gate to go with our current IPv6 API tests. 20:11:25 <cgoncalves_> that would be job octavia-v2-dsvm-scenario-ipv6 for IPv6-only deployments 20:11:30 <johnsom> The multi-node gate work is stalled for a bit. I got stuck so stopped to focus on IPv6. Currently keystone isn't working right. I think it's due to the HOST_IP setting for the multinode setup, but still need to dig on that. 20:12:11 <johnsom> Well, I would add it to our existing tests like we do for API. I think the only reason we would need another job is for IPv6 control plane testing. 20:12:35 <cgoncalves_> correct 20:13:11 <johnsom> Otherwise I also need to work on setting up octavia-lib. We have the repo, it's just empty at the moment. 20:13:33 <johnsom> Any other updates from folks? 20:13:56 <johnsom> BTW, if you have cycles, I would welcome another set of eyes on the multi-node stuff. 20:13:58 <cgoncalves_> as folks might have noticed, job octavia-v2-dsvm-scenario-centos-7 keeps on failing intermittently 20:14:35 <johnsom> Yes, it has been. I am getting close to marking it non-voting. 20:14:42 <cgoncalves_> we thought we had fixed it a few months ago but while it seemed to have improved success rate, it kept on failing from time to time 20:15:04 <rm_work> yeah, i'd say go to non-voting for a bit, it's been a little wacky 20:15:16 <rm_work> i wish i had time to look at that... 20:15:17 <cgoncalves_> thanks to johnsom's work on using CI mirrors and this patch from Ian, the job should be good soon 20:15:18 <rm_work> or anything... 20:15:19 <cgoncalves_> #link https://review.openstack.org/#/c/609169/ 20:15:51 <johnsom> Yep, Ian and cgoncalves have been looking at it, so I have hope it will improve soon. 20:16:25 <cgoncalves_> it seems to work fine. we could do my rechecks on https://review.openstack.org/#/c/609170/ to make sure but I'd say we're fine now 20:16:58 <johnsom> I will take a look at his test job output and vote on the DIB patch 20:17:08 <cgoncalves_> thanks 20:17:55 <johnsom> I have also been helping one of the vendors. They have had some questions. I think they are pretty close with a driver. So, that is good. 20:18:08 <johnsom> I am probably the part holding them up with the octavia-lib work 20:18:44 <cgoncalves_> awesome 20:19:34 <johnsom> Any other updates? 20:19:56 <rm_work> yeah will be great to finally say we have providers implemented and actually have some third-party ones functioning 20:20:44 <johnsom> Yep 20:21:02 <cgoncalves_> the OVN provider folks have also been participating with some work in octavia-tempest-plugin 20:21:44 <johnsom> Yeah, I know Reedip was working on adding the "disable features" options. I need to go look at how that is coming along. 20:21:48 <cgoncalves_> we should probably spend some time and have it merged. it will allow us to add another open-source provider driver to the list 20:22:14 <johnsom> I also noticed the OVN driver is being bad and reaching into our code to go talk to neutron....... 20:23:16 <johnsom> #topic Open Discussion 20:23:18 <cgoncalves_> I believe they have a to-do list of things to fix and improve. 20:23:41 <johnsom> Ok, moving on to other topics. Anyone have other topics for today? 20:24:21 <johnsom> FYI, I put up a patch that adds the API versioning to the api-ref: 20:24:23 <johnsom> #link https://review.openstack.org/604911 20:24:46 <johnsom> This is the IPv6 patch I have been working on: 20:24:48 <johnsom> #link https://review.openstack.org/589292 20:24:57 <cgoncalves_> I would like to have https://review.openstack.org/#/q/ad69363fc753c0689214fcdc26d6ef6c90d0d520 backported. I sent an email to the openstack-dev mailer because I had some concerns/doubts about the stable policy 20:24:59 <cgoncalves_> #link http://lists.openstack.org/pipermail/openstack-dev/2018-October/135530.html 20:25:00 <johnsom> If you want to give me comments while I'm still working on it. 20:25:11 <johnsom> I have one issue left to resolve 20:25:19 <cgoncalves_> I still need to find some time to go and address Matt's comments 20:25:34 <johnsom> Yeah, I saw the email chain 20:26:22 <cgoncalves_> he made a good point, the release note has a 'feature' item that should have been under 'fixes' 20:26:58 <cgoncalves_> johnsom: oh, that api-ref patch, thanks! 20:27:32 <cgoncalves_> that was something someone asked the other day 20:27:48 <johnsom> Yeah, oversight on our part. 20:28:10 <cgoncalves_> we learn as we go 20:28:32 <johnsom> Yep, as long as we are making incremental progress.... 20:28:57 <johnsom> Well, if no one has other items to discuss, I will get back to working on the IPv6 stuffs 20:29:34 <johnsom> #endmeeting