18:00:10 #startmeeting networking_ovn 18:00:11 Meeting started Thu Sep 8 18:00:10 2016 UTC and is due to finish in 60 minutes. The chair is rtheis. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:12 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:00:14 The meeting name has been set to 'networking_ovn' 18:00:19 mmm, official 18:00:25 Hi 18:00:28 i was about to ask, first topic, which meeting topic to use 18:00:29 Hello 18:00:33 could have also been lazy and used "ovn" 18:00:50 networking-ovn makes sense 18:00:51 Hello! 18:00:54 hi! 18:00:54 since thats the project name 18:00:58 yep, it's fine 18:01:04 yeah, I think networking_ovn was in the meetings patch 18:01:07 but it needs to be consistent for people looking at logs 18:01:16 rtheis: cool, sounds good. 18:01:17 welcome to the first official meeting 18:01:23 thanks for setting it up, rtheis 18:01:29 we've outgrown the main OVN meeting at this point 18:01:44 yw 18:02:04 rtheis: have an agenda in mind? 18:02:04 #topic Announcements 18:02:11 #link https://etherpad.openstack.org/p/networking-ovn-meeting 18:02:27 I've placed the proposed agenda here 18:02:35 nice. 18:02:42 The Newton-3 deliverable was released 18:02:48 #link https://github.com/openstack/releases/commit/93d18e485ad2043df9a793dc2a6cd580d999bd0d 18:03:04 next week is Newton RC1 18:03:16 #link https://releases.openstack.org/newton/schedule.html 18:03:38 I am planning to propose a patch for Newton RC1 next week 18:04:13 should we merge all feature freeze exceptions by then? 18:04:34 That's what I'd like to discuss 18:04:49 We have several bug fixes and a lot of features in the review queue 18:05:09 and would like opinions how what to let into Newton RC1 18:05:26 as much as possible? :) 18:05:37 sounds good :) 18:05:47 prioritizing makes sense though 18:05:57 yeah 18:06:06 any other announcement before we move on ? 18:06:27 #topic Bugs for Newton RC1 18:06:28 i mentioned this in the main OVN meeting, but TripleO support is now all merged 18:06:43 nice 18:07:07 are we going to provide distro-package install instructions anywhere? 18:07:10 speaking of rh 18:07:15 i dont know if ubuntu is packaging ovn 18:07:22 Sam-I-Am: we'll have TripleO docs at some point 18:07:31 our general install guide mentions packages that are available 18:07:40 have they been tested? 18:07:42 i would hope that all distros start turning on the OVN sub-package in OVS 2.6 18:07:50 the packaging work was done long ago 18:07:53 it would be nice to make sure that stuff works 18:07:59 might have to ping the ubuntu ovs package maintainer though 18:08:01 not sure who that is 18:08:21 probably the openstack packagers 18:08:31 at least thats how its been for the cloud-archive stuff 18:09:02 any work on the networking-ovn side that we need to track for this? 18:09:12 bug for RC1? 18:09:17 ubuntu needs to package networking-ovn 18:09:24 if anyone wants to chase that ... 18:10:00 i suppose i can poke at that 18:10:03 it's already packaged for centos / rdo 18:10:05 i know people 18:10:09 cool. 18:10:12 thanks Sam-I-Am 18:10:20 not that they're terribly responsive... but we shall see 18:10:36 anything else on packaging? 18:10:43 having package options for both rdo and ubuntu makes sense for our docs 18:10:48 since we more or less have two camps 18:11:17 #link https://bugs.launchpad.net/networking-ovn/+bug/1514488 18:11:18 Launchpad bug 1514488 in networking-ovn "Make use of the native OVN DHCPv4" [High,Fix released] - Assigned to Numan Siddique (numansiddique) 18:11:46 We have several fixes related to native DHCP support listed in agenda 18:12:00 I think we should try targeting them for RC1 18:12:29 lets just not release anything thats broken :) 18:12:47 +1 18:12:49 numan is on vacation for 3 weeks 18:12:54 rtheis: is there a reason why the Vagrant in OVN (networking-ovn) is not using the native DHCP? Maybe metadata issues? 18:12:57 who wrote that originally, so we'll have to proceed on those fixes without him 18:13:12 Zong Kia has picked up a couple 18:13:31 flaviof: just haven't made the switch yet 18:13:48 probably just that not everyone uses the vagrant setup 18:13:53 conventional md doesnt work at all with native dhcp? 18:14:03 no 18:14:10 ack. so it is just not a known issue, but yet not fully tested then. 18:14:25 thus, i'd say native dhcp is not ready. 18:14:36 understood. 18:14:47 people expect basic features to work 18:15:06 so perhaps native dhcp is still 'experimental' and not a default? 18:15:08 config drive is the alternative which does work 18:15:23 with the native services 18:15:30 it is, but not everyone uses it 18:16:48 just to be clear, I'm talking about https://review.openstack.org/#/c/350237/7/vagrant/provisioning/setup-compute.sh -- line 36 18:17:03 the metadata limitation and other limitations for native services are documented in the release note 18:17:37 #link http://docs.openstack.org/releasenotes/networking-ovn/unreleased.html 18:18:13 make sense. all good. 18:18:33 flaviof: the vagrant setup was setting up both native and conventional dhcp 18:18:49 I fixed it to go back to conventional dhcp 18:18:55 heh, ic; that is defininitely bad 18:18:57 just haven't got back to switching it to native 18:19:20 since there was documentation and other things to do 18:19:49 any other concerns with native DHCP fixes ? 18:20:33 looks like ubuntu is including a networking-ovn package 18:20:34 none from me; thank y'all for taking the cycles to educate me on config-drive 18:20:41 Sam-I-Am: great! thanks for checking. 18:20:48 but i'm not sure if they're depending on a useful ovs version or kernel 18:20:53 so whether it works or not remains to be seen 18:21:12 #link https://bugs.launchpad.net/networking-ovn/+bug/1611963 18:21:13 Launchpad bug 1611963 in networking-ovn "OVN NB sync ignores ACLs on lswitch without ports" [Undecided,In progress] - Assigned to Richard Theis (rtheis) 18:21:21 #link https://bugs.launchpad.net/networking-ovn/+bug/1612435 18:21:22 Launchpad bug 1612435 in networking-ovn "ML2 driver doesn't retry post_fork_initialize ovn nb or sb connection failures" [Undecided,In progress] - Assigned to Richard Theis (rtheis) 18:21:41 there are a couple miscellaneous fixes for these bugs 18:21:58 sync and connection retry 18:22:08 any concerns with targeting RC1 ? 18:22:48 also ... 18:22:50 #link https://review.openstack.org/#/c/358501/ 18:23:46 #topic Feature Freeze Exceptions (FFEs) for Newton RC1 18:24:01 We have 7 listed in the agenda 18:24:31 This is a lot of new function potentially coming in late to the release 18:24:52 are there any features that should be blocked? 18:25:34 or maybe we focus on which feature is the highest priority and work down from there 18:25:43 i think we should prioritize 18:25:52 one of my top priorities would be https://review.openstack.org/#/c/346646/ 18:25:58 is chandra working on the test failure? 18:26:13 next for me would be vlan-aware-vms 18:26:14 yes 18:26:35 I believe chandra has an ovn patch dependency 18:26:43 ah, i see 18:26:52 which what I think he brought up in ovn meeting 18:26:53 maybe that's the patch he mentioned in the ovn meeting 18:26:55 got it. 18:26:57 yeah 18:28:03 anyone else have opinions on priority ? 18:28:36 Is race condition a concern? 18:29:06 zhouhan: is that for https://review.openstack.org/#/c/358447/ ? 18:29:08 I am working on the patch but not sure about priority 18:31:15 priority may depend on how easy it is to recreate the race conditions in production 18:31:15 What about the idl mutate patch: https://review.openstack.org/#/c/351861/ 18:31:19 zhouhan: i feel like that might be too invasive for this release 18:31:22 I think I just got dropped 18:31:35 I tend to agree with russellb 18:32:14 ovs mutate -- what perf improvement do we get? 18:32:29 measured that? just curious 18:32:51 https://bugs.launchpad.net/networking-ovn/+bug/1610347/comments/2 18:32:53 Launchpad bug 1610347 in networking-ovn "OVS transaction timeout with many logical switch ports on a logical switch" [Undecided,In progress] - Assigned to Richard Theis (rtheis) 18:33:14 That's a question, I don't have an answer to yet - our internal cloud where we going to test this out has been having stability issues lately 18:33:27 rtheis: nice, thanks 18:33:29 I noted this improvement in a local environment 18:33:30 We did get past our initial problem point. 18:34:23 russellb: well, it is not that invasive as I thought in the beginning. I didn't change the sync call behavior, not like the journal mechanism in ODL 18:34:43 zhouhan_: oh ok, interesting 18:35:00 rtheis: and https://review.openstack.org/#/c/362494/ 18:35:19 zhouhan_ thx 18:35:58 It is POC so I'd like to get feedback if that is a good way to continue 18:36:21 zhouhan_: the tests aren't passing on these 2 patches yet... do you think we give them more time for reviews and testing and possibly backport if not to invasive? 18:37:05 rtheis: yes, I am fixing the failure and added some functional tests for race conditions 18:37:12 ack 18:37:24 It is not quite urgent though. we know that, race conditions ... 18:37:44 so our priority so far is ... 18:37:47 (1) NAT support (SNAT, FloatingIP) 18:37:54 (2) OVN trunk driver to support vlan-aware-vms 18:38:03 (3) Add OVS transaction mutate support 18:38:21 (4) multi-provider / multi-segment 18:39:25 should 3) be higher than 2)? 18:39:32 @rtheis: is vlan-aware VMs the way to provide services ? 18:39:55 ramamurthy: I haven't looked at that patch yet 18:39:58 russellb ? 18:40:20 For my understanding 3) is about scalability and performance, but not sure if it really solves the problem 18:41:17 swapping 2 and 3 is fine with me 18:41:24 i actually want both :) 18:41:27 so whichever :) 18:41:56 i'd rather defer the metadata one, i think ... 18:42:26 zhouhan_: it did seem to solve the problem for the use case reported 18:42:36 which problem did it not solve? 18:43:00 rtheis: that's great. I didn't test, so not sure 18:43:06 ack 18:43:34 if we defer metadata then we have ... 18:43:43 Support a mixed DPDK and non-DPDK environment 18:43:51 Support native OVN DHCPv6 18:43:56 leve 18:43:57 left 18:44:00 DPDK one is pretty small 18:44:08 so gets my +1 18:44:20 sounds good 18:44:25 is anyone here actively involved with SFC (specifically, SFC with OVN) ? 18:44:39 and of course, would love DHCPv6 too, but looked like there was still work needed? 18:44:48 i'd say (5) dpdk, (6) dhcpv6 18:44:58 and then try to merge all 6 :) 18:45:09 awesome :) 18:45:25 Zong Kai picked up DHCPv6 while numan is out 18:45:31 great 18:45:34 we'll see how far he gets 18:45:48 i'm on vacation all next week ... and technically in meetings today and tomorrow 18:45:50 how well does ipv6 work in general? 18:45:51 but will do as much as i can 18:47:07 mighty quiet :) 18:47:37 russellb: is ovn ipv6 support ready ? 18:47:51 yes 18:47:56 sorry in meeting room 18:48:06 np 18:48:13 IPv6 support is merged in OVN for L3 routing 18:48:20 there are improvements still to be made 18:48:30 ok 18:48:53 so we have a list of 1-6, with metadata (7) deferred 18:49:00 like, i don't think we're doing router advertisements? 18:49:13 for our logical routers, i mean 18:50:23 thats sort of a problem 18:50:24 anything else for Newton RC1 that we want/need to complete? 18:51:28 all the things 18:51:40 rtheis: thanks for organizing the agenda 18:51:51 you're welcome 18:51:56 last topic ... 18:51:58 #topic Open Discussion 18:52:10 anything else for today ? 18:52:14 is anyone here actively involved with SFC (specifically, SFC with OVN) ? 18:52:24 I am not 18:52:28 i've monitored it somewhat, not too up to date 18:52:34 ryan was commenting a bunch a while back 18:52:42 there is a networking-sfc spec i was asked to review in the last couple weeks 18:53:00 but the main issue with it is it makes assumptions / proposals for the SFC support in OVN itself, which really needs to get done first 18:53:02 #link https://review.openstack.org/#/c/333172/ 18:53:15 ack. thanks! 18:53:28 i haven't seen a proposal for OVN itself in a while 18:53:33 or at all? lots of discussion 18:53:55 but i'd like to see a specific design proposal 18:53:57 yeah, I'm catching up on all that was talked about in the ML 18:54:45 rtheis: thanks for the link! 18:54:49 yw 18:55:07 the ovn discussion didnt focus on specific usecases 18:55:12 for SFC 18:55:59 Ryan mentioned something about it being single tenant only - but I've not looked into it 18:56:07 right. there is the open topic on multi-tenants, as an example. let me educate myself more on that b4 I start making my usual stupid questions :) 18:56:45 any other topics for today? we have a couple minutes left. 18:57:22 chandrav: I know you just joined, your SNAT/FIP patch was voted as top priority for review :) 18:57:34 let's plan an OVN meetup in Barcelona for whoever is there 18:57:34 another related event for this group: OVSCon details were just sent out 18:57:34 accepting talk proposals now 18:57:38 i still don't understand the single vs multi tenant concern 18:57:56 azbiswas: Thats good to know :) 18:58:01 yep, just needs tests to pass 18:58:03 then let's get it in 18:58:09 OVN meetup sounds good 18:58:13 chandrav: you need an OVN patch in first? 18:58:14 thanks chandrav 18:58:17 russellb: tbh I don't either, I'm just repeating what I heard on that as a parrot at this point 18:58:36 for the tests to pass, i'll point to my ovs branch 18:58:38 chandrav: do you have a link to the patch ? 18:59:13 https://patchwork.ozlabs.org/patch/665301/ 18:59:23 #link https://patchwork.ozlabs.org/patch/665301/ 18:59:24 thanks 18:59:26 russellb: sorry that I may not go to Barcelona, you guys have fun 18:59:43 I am having difficulty writing a unit test case for that patch. 18:59:44 zhouhan_: how about ovscon? 18:59:49 at least that one is in your back yard :) 19:00:03 russellb: I hope I can join that one :) 19:00:11 zhouhan_: +1 19:00:24 we are out of time, thank you all for attending 19:00:34 thanks guys, bye 19:00:36 ##endmeeting 19:00:41 #endmeeting