15:01:36 #startmeeting neutron_l3 15:01:37 Meeting started Thu Sep 4 15:01:36 2014 UTC and is due to finish in 60 minutes. The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:01:39 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:40 #topic Announcements 15:01:41 The meeting name has been set to 'neutron_l3' 15:01:55 #link https://wiki.openstack.org/wiki/Meetings/Neutron-L3-Subteam 15:02:34 I’d like a short meeting today. 15:02:59 Juno-3 should be available today. I know the cut-offs are always difficult. 15:03:36 I did not get a chance to update the agenda on the wiki. I don’t know where the week went (actually I do, I took two days off.). 15:03:50 Any other announcements? 15:04:04 nothing from me 15:04:22 #topic neutron-ovs-dvr 15:04:49 Swami: mrsmith: viveknarasimhan: armax: What is the latest on DVR? 15:06:45 the backlog was getting down and has creeped up a little bit in the last week or two. It still is pretty much under control though. 15:07:08 I think the reason for the creep was that we had our attention diverted to getting new features in to juno-3. 15:07:18 i wonder if there's much point to discuss dvr progress here while it has its own dedicated meeting 15:07:56 yamamoto_: We’ve needed it in the past but now that it is slowing and stabilizing, you may be right. 15:08:06 real quick - I am focused on the migration patch and the newly split up snat fixes patch 15:08:13 to me those are priorities 15:08:27 yamamoto_: Still, we report progress up through the L3 team so it is good to touch on it. 15:08:54 carl_baldwin: fine. just wondering 15:08:56 the migration patch depends on the snat fixes so hopefully we will get those thru soon 15:09:14 mrsmith: Right. I think I will take my co-author line off of the snat patches so that I can +2 them. I’ve been thruogh them pretty thoroughly. 15:09:20 depends = it needs the fixes to fully function 15:09:37 carl_baldwin: gotcha 15:09:58 mrsmith: understood. I will be hitting DVR bugs again as a priority (only behind wrapping up L3 HA). 15:10:17 carl_baldwin: isn't it better to find another core reviewer if you are actually a co-author? 15:10:56 I believe carl mostly just split up the big patch into smaller ones 15:10:59 yamamoto_: I was only an author in the sense that I split the original patch up and did very minor fixups. It was a bit of a stretch to add the co-author line, I’ll admit. 15:11:32 carl_baldwin: ok it makes sense. thank you 15:11:50 yamamoto_: You’re right though. Credit should be given where it is due. 15:12:39 On the other hand, “It is amazing what you can accomplish if you do not care who gets the credit.” - Harry S. Truman (I think) 15:12:40 :) 15:12:54 Anyway, back to the meeting. 15:12:57 Anything more on DVR? 15:13:31 nothing from me 15:13:45 #topic l3-high-availability 15:13:59 amuller: Do you have a report? 15:14:09 Sylvain is addressing comments on the server side patches 15:14:19 Welcome back, safchain 15:14:23 hi 15:14:25 I'm working on the agent patches, improving our functional tesitng 15:14:28 testing* 15:14:49 I'm addressing comments about the vr_ids management 15:15:05 Maru and John Schwarz have been dealing with the inconsistent failures at the gate, and have already fixed it, but it'll take a while to merge because of the gate 15:15:09 safchain: Which patch? I didn’t notice the comments. 15:15:22 base classes 15:15:33 around the _set_vr_id method 15:16:43 carl_baldwin, https://review.openstack.org/#/c/64553 15:16:46 safchain: right. I see it. 15:17:37 That is good feedback. 15:17:48 carl_baldwin, so I introduced a new table for the allocations 15:18:01 Darragh O'Reilly reported an issue in the keepalived.conf after we add FIPs, trying to reproduce and test it 15:18:01 I mean vr_id allocation per ha network 15:19:16 safchain: Feel free to ping me if you need anything or when you have something ready for review. 15:19:19 Also we have a bug tag now and a bunch of bugs to go with: https://bugs.launchpad.net/neutron/+bugs?field.tag=l3-ha 15:19:29 some are closer to RFE 15:19:31 carl_baldwin, ok, thx 15:20:21 amuller: I will start to work on triage and assigning importance. 15:20:42 amuller: Thanks for gathering them together and posting the query. 15:21:07 carl_baldwin: the feature is targeted to RC1 15:21:16 when do we expect to merge the original 3 feature patches? 15:21:27 RC1 is too late imo 15:21:37 we need to deal with stability, bugs, testing 15:22:07 RC1 is on the 25th 15:22:22 amuller: I agree. This is now my top priority to turn these reviews around. 15:22:59 alright 15:23:06 I feel the base classes patch is in pretty decent shape. 15:23:10 now that Sylvain is back it should go faster too 15:23:27 For the other two, I made my comments and I don’t feel that there are any hard issues in there from my perspective. 15:23:30 amuller, yes I hope 15:24:17 Please ping me as soon as new patches are ready for review so that we can turn them around quickly. 15:24:23 Sounds good 15:24:31 amuller: anything else? 15:24:49 working with Maru on the agent side failover functional test 15:25:05 #action carl_baldwin will triage L3 HA bugs 15:25:11 I think it's pretty cool, it's a step to having multiple L3 agents on a single machine 15:25:16 so we can test this in Tempest as well 15:25:30 also we've been talking about integration testing in-tree 15:25:41 amuller: That sounds cool. 15:25:50 if we can get multiple L3 agents up it'll help with that as well, should help DVR and L2pop as well 15:25:58 other features that require multiple agents 15:26:21 amuller, carl_baldwin yes that is really something that we need 15:26:38 I hope you can raise the bar for testing in the project. 15:26:53 it's all marun_ guiding things 15:27:33 Aiming for the failover testing in the next week-ish, and in-tree integration testing probably for early Kilo 15:27:44 That's it for me 15:28:18 amuller: Great, thanks for the update. 15:28:36 #topic bgp-dynamic-routing 15:29:11 was it deferred to kilo? 15:29:11 devvesa: ping 15:29:18 yamamoto_: it was 15:29:30 why? 15:30:08 yamamoto_: Reviewer bandwidth, really. 15:30:48 I guess it was more than that. Juno had a lot of work done in it and we can only pack so much in. 15:31:47 This will be our first priority to open up Kilo development. 15:32:12 is it a candidate of neutron-incubator? 15:32:49 yamamoto_: It was a candidate but I’ll admit that I’ve fallen a bit behind in the neutron incubator discussion. 15:33:50 i just hope neutron-incubator stuff doesn't defer this further. 15:34:08 it really depends on the community discussion. we also need to consider L3 is tightly coupled with L2 and many folks regard it as a part of core. 15:34:55 yamamoto_: If neutron-incubator is not ready, I will continue with dynamic routing in the main repository. 15:35:57 carl_baldwin: it makes sense 15:36:40 #topic Kilo 15:37:48 I don’t have much time to get in to it today but we should probably have Kilo in the back of our minds and start discussing what we would like the main themes to be. So, I’m going to add a Kilo topic to the weekly agenda. 15:38:22 The design sessions are still not open, http://summit.openstack.org/ 15:38:47 Redesign the L3 agent, since we added distributed and HA routers it's an unmaintainable mess 15:38:51 seizadi: when it will? 15:38:55 I think that's a strong candidate for early Kilo 15:39:20 amuller: +1 refactor l3-agent 15:39:22 amuller: +1 15:39:43 amuller: +1, I have had some refactors in mind myself. 15:39:54 structure it somehow, add more classes, enough with these 'if distributed, if not distributed, if ha, if not ha' nonsense 15:40:12 Hopefully we can have a good idea of where to start when Kilo opens. 15:40:13 amuller: +1. I hope it includes vpn-agent merging into l3-agent. 15:40:44 amuller: Exactly. 15:40:49 +1 15:41:09 amotoki: vpn merging noted 15:42:15 splitting up process_router() is another candidate 15:42:19 it is getting huge 15:42:25 we could start with that yeah 15:42:32 yamamoto: I will post when I see it open, I thought if Carl finds out design summit schedule for Kilo through Openstack communication, we can discuss on this topic. 15:42:49 mrsmith: +1 15:43:19 The ML will probably be the best place to watch for it to open. 15:43:30 mrsmith: +1 15:43:43 It can’t be too much longer. We’ve only got two months until the summit. 15:44:08 #topic Open Discussion 15:44:22 Anything else? 15:44:34 nothing from me 15:44:39 one question on relationship between DVR and L3-HA. 15:44:49 Are they exclusive? just confirmation. 15:44:53 for now, yes 15:45:05 we just didn't get to it yet 15:45:11 #link https://bugs.launchpad.net/neutron/+bug/1365473 15:45:33 I am now planning to write a small patch in horizon and I need some constraint check for them. 15:45:47 *small patch to support L3-HA 15:45:54 next patchset for the models patch should block the creation of a rotuer that's both HA and distributed 15:45:57 router* 15:46:00 If anyone is on the IPv6 subcommittee I like to talk to them we are having trouble with IPv6 integration. I looked at their meeting notes but not much notes. 15:46:48 rajeev wanted me to ask for reviewers on a couple of his patches 15:47:01 https://review.openstack.org/#/c/116412/ 15:47:29 https://review.openstack.org/#/c/112146/23 15:48:03 i think sc68cal is the right contact. 15:48:15 mrsmith: noted 15:49:12 Okay, my time is up. I’ll be back online in a bit. 15:49:15 #endmeeting