15:00:10 #startmeeting neutron_l3 15:00:10 Meeting started Thu Jul 16 15:00:10 2015 UTC and is due to finish in 60 minutes. The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:10 hi 15:00:11 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:14 The meeting name has been set to 'neutron_l3' 15:00:15 hi 15:00:19 morning 15:00:22 #topic Announcements 15:01:02 I bet Liberty-2 is coming up soon. Let me see... 15:01:14 #link https://wiki.openstack.org/wiki/Liberty_Release_Schedule 15:01:26 Looks like we have — at most — two weeks. 15:01:39 Any other announcements? 15:01:45 Oh, 15:01:52 #link https://wiki.openstack.org/wiki/Meetings/Neutron-L3-Subteam 15:02:16 Any other announcements? 15:02:17 hello 15:02:23 under "apologies for absence" - I've got another meeting at the bottom of the hour, so I'll have to do readback of the second half and followup on ML or in channel 15:02:46 regXboi: ack 15:02:58 #topic Bugs 15:03:00 mlavalle: ping 15:03:05 hi 15:03:30 Fix committed for https://bugs.launchpad.net/neutron/+bug/1465434 15:03:30 Launchpad bug 1465434 in neutron "DVR issues with supporting multiple subnets per network on DVR routers - PortbindingContext does not have the status." [Critical,Fix committed] - Assigned to Swaminathan Vasudevan (swaminathan-vasudevan) 15:03:40 Great work Swami 15:04:03 Also had a DVR regression sneak in but I think that should be fixed (or soon) 15:04:48 haleyb: Do you have a link, by chance, to the review? 15:04:58 * carl_baldwin can’t find it at the moment. 15:05:03 https://review.openstack.org/#/c/201907/ 15:05:12 merged this morning 15:05:20 haleyb: thanks, excellent. 15:05:26 there was a follow-up 15:05:45 https://review.openstack.org/#/c/202274/ 15:05:52 that needs an update 15:06:11 dvr job was clean on 201907. Let’s keep an eye on it and get the job voting soon. 15:06:46 #action carl_baldwin will follow up with Swami on 202274 15:06:55 Thanks, haleyb 15:07:01 Any other bugs to discuss? 15:07:26 np, it does remind me to be cautious of the aggressive mocking done in unit tests 15:07:28 carl_baldwin: a fix was merged for https://bugs.launchpad.net/neutron/+bug/1323766. I had a conversation with dustin last week and concluded it was different to 15:07:28 Launchpad bug 1323766 in neutron "Incorrect Floating IP behavior in dual stack or ipv6 only network" [Medium,Fix committed] - Assigned to Dustin Lundquist (dlundquist) 15:08:00 https://bugs.launchpad.net/neutron/+bug/1469322 15:08:00 Launchpad bug 1469322 in neutron "Associating a floatingip with a dual stack port requires the fixed-address to be specified" [High,Fix committed] - Assigned to Dustin Lundquist (dlundquist) 15:08:02 mlavalle: Thanks for following up on that. 15:08:15 this last one is still in progress 15:08:22 carl_baldwin: Will try to remember to ping amototoki about https://bugs.launchpad.net/neutron/+bug/1471316 later today 15:08:22 Launchpad bug 1471316 in neutron "_get_subnetpool_id does not return None when a cidr is specified and a subnetpool_id isn't." [High,In progress] - Assigned to John Davidge (john-davidge) 15:08:45 but at least we are not confused with the similarities of those two bugs 15:09:04 john-davidge: Thanks, I’ve been in a holding pattern on that one. It will help to ping amotoki 15:10:02 carl_baldwin: the one that swami fixed is the same amuller was inquiring about last week, right? 15:10:04 john-davidge: I’ll send him email too. 15:10:18 mlavalle: Need more context... 15:10:32 carl_baldwin: Great, thanks. I’m going to be in meetings for the rest of the day so that would be a great help 15:10:34 #action carl_baldwin will ping amotoki on 1471316 15:11:00 carl_baldwin: nm... i answerd my question myself.... :-) 15:11:30 mlavalle: ok 15:11:56 #topic Routed networks 15:12:41 Not much here. markmcclain has provided some feedback and caused me to rethink some things. I will digest and send mail to the ML. 15:13:42 I’ve been thinking this for a while but there really isn’t much chance for anything making Liberty on this subject. This will be a priority for Mitaka (did I get the name right?). 15:13:52 yes - that's the right name 15:14:16 * regXboi double-checks his mailboxes 15:14:19 #topic BGP dynamic routing 15:15:04 hi 15:15:05 * carl_baldwin curious if the Japanese characters will render correctly in everyone’s IRC. 15:15:10 hi 15:15:11 tidwellr: hi 15:15:13 carl_baldwin: is there any chance (if additional resources were found) of getting bgp listeners and RIB in as part of this in Liberty? 15:15:28 doubtful 15:15:43 * regXboi sighs and says "ok" 15:15:44 ok 15:15:51 I think someone needs to lay out a use case for it 15:15:55 regXboi: Honestly, resources won’t help at this point. We’re almost to Liberty-2 and we’d have to go through design. 15:16:16 mythical man-month and all. 15:16:24 tidwellr: I think use cases are pretty easy to come by 15:16:33 carl_baldwin: that was the thing worrying me 15:16:40 regXboi: It will make a good start. 15:16:48 ok, maybe start an RFE and take the discussion there? 15:16:48 Hi there, sorry for coming in late... 15:16:55 tidwellr: Vikram: How is the rest going? 15:17:01 ok, give me the action to lay out use cases for listener 15:17:13 on an etherpad 15:17:15 Any chance we can have functional code by around Liberty-2? 15:17:42 regXboi: I encourage people to add their own actions. It is allowed by the meeting bot. 15:18:04 carl_baldwin: ah - did not realize that 15:18:08 I'm proposing some changes that should simplify things and help us go faster 15:18:24 Carl: Me and Ryan working towards the agent.. 15:18:32 #action regXboi to start etherpad for collecting use cases for implementing bgp listener and RIB 15:18:56 the biggest things outstanding are really the agent the DB queries that scoop all the routes we need to advertise 15:18:57 It's good Numan will also help us for this project, so i except the code should be ready soon 15:19:52 I think we have a good head start on the agent, the DB queries are lots of fun ;) and I'm working through those 15:20:12 Just as a time check. It will likely take a few weeks for review and merging code. And the code must merge for Liberty-3. Keep that it mind. 15:20:19 Liberty-2 is a noble goal, not too far out of reach 15:20:20 * mlavalle also having fun with db queries 15:20:21 tidwellr: is there an early draft of the DB queries? 15:20:30 tidwellr: I have also started the work for driver. Patch will be out soon 15:20:36 * regXboi has some experience crafting queries 15:21:09 Vikram: thanks. Be sure to ping me and others who are interested when there is something to review. I gave up reading gerrit notifications long ago (too high volume). 15:21:25 will the agent query the db? 15:21:37 carl_baldwin: ok 15:21:38 queries are server-side 15:22:08 tidwellr: gotcha 15:22:19 Anything more to cover here? 15:22:37 tidwellr, mlavelle - if either of you need help with crafting, i'd be willing to spend some cycles 15:22:43 regXboi: SQL and sqlalchemy are not the trick, it's navigating the neutron schema 15:22:54 tidwellr: understood 15:23:13 some pointers are always welcome 15:23:17 regXboi: I'll ping you if I get stuck thanks :-) 15:23:35 regXboi: see https://review.openstack.org/#/c/201621/4/neutron/db/bgp_speaker_db.py for the first pass at what I've got 15:23:44 tidwellr: ok, will look 15:23:54 and will comment tere 15:23:55 er there 15:24:13 I know that query is broken, but it's a start 15:24:22 sure 15:25:19 Are we good on BGP? 15:25:38 looks like it 15:25:40 #topic IPAM 15:25:50 pavel_bondar: johnbelamaric: hi 15:25:57 carl_baldwin: hi 15:25:59 carl_baldwin: hi 15:26:16 pavel_bondar has updated #link https://review.openstack.org/#/c/183099/ 15:26:16 One more patch to go. 15:26:22 i am reviewing it now 15:26:37 *crossing his fingers* 15:26:43 I have added some more refactoring and did additional cleanup 15:26:56 Ok, I plan to give it another go tomorrow. I tried getting through it, just ran out of time. 15:27:13 Ok, thanks 15:27:35 Or, maybe today. johnbelamaric, ping me when you’ve been through it. 15:27:47 carl_baldwin: will do - today would be great :) 15:27:58 I don’t want to delay it if I can help it. 15:28:21 carl_baldwin: thanks, it's appreciated 15:28:28 carl_baldwin: thanks 15:28:30 Is there any follow-up to consider after this patch merges? 15:29:19 carl_baldwin: not that I know of immediately. there were some items along the way. pavel - how are the CI's looking with it enabled by default 15:29:37 carl_baldwin: do we need some kind of migration to IPAM? 15:30:10 carl_baldwin: earlier we considered only green fields deployment will use pluggable ipam 15:30:24 pavel_bondar: that will be needed if we were to make it default. it should be added at that time or before 15:30:44 pavel_bondar: but that would just be for reference driver of course. alternate driver authors will need their own migrations 15:30:47 are we planning on making it default in Liberty? 15:30:53 regXboi: no 15:30:55 pavel_bondar: johnbelamaric: What migration is needed to switch to pluggable reference implementation? 15:31:16 carl_baldwin: you need to populate the reference driver's tables with the subnet info, etc. 15:31:18 johnbelamaric: thx, I didn't think so, so I was getting nervous 15:31:53 carl_baldwin: the ref driver stores it in its own private area, so that it operates like an external driver 15:32:09 johnbelamaric: ok. We should be thinking about that for early Mitaka 15:32:24 carl_baldwin: makes sense 15:32:34 carl_baldwin: sound good to me, thanks 15:32:53 Anything else on IPAM? 15:33:08 no from my side 15:33:11 nope, that's it i think 15:33:30 #topic ML3 router plugin 15:35:28 We’ll probably save this one for later in the interest of time. 15:35:39 #topic DNS 15:35:43 mlavalle: hi 15:35:58 As you know, I've pushed wip code for the internal side 15:36:02 It was good to start reviewing yesterday. 15:36:23 I got reviews from you and others, so i know i didn't go off the rails ;-) 15:36:39 my goal is to have this code functional by the end of this week 15:36:53 mlavalle: That is great. 15:36:56 also starting this week the external side 15:37:22 and the nova side we are going to address with a FFE 15:37:36 * mlavalle wonders if that is the right term 15:37:54 so working to have this done in Liberty 15:38:06 My concern is mostly around the new model. 15:38:11 #link https://review.openstack.org/#/c/200952/3/neutron/db/models_v2.py 15:38:23 * mlavalle listening 15:38:53 you want to discuss here or is it already in gerrit? 15:39:19 I had imagined just a dns_name attached to a port. Either directly or with a table with port_id and dns_name as columns. 15:39:56 The model has much more. We don’t need to discuss it here but I wanted to bring it up so others like HenryG and Ann (don’t know nick) could go look. 15:40:17 i think it might be simplified. I just used the IPAllocation table as a guide...... 15:40:49 but yeah, let's discuss it in gerrit 15:41:06 thanks for the comments... that is what i needed at this point 15:41:12 mlavalle: Let’s iterate on it a bit in the next couple of days. I think we can get it simplified but maybe there is good reason for a little more than what I had imagined. 15:41:12 :-) 15:41:25 mlavalle: Keep up the good work. 15:41:33 will do 15:41:40 that's all i have today 15:41:47 #topic Address Scopes 15:42:18 I have this working locally. I have been trying to get patches in to shape and spoon feeding them to gerrit. 15:43:14 #link https://review.openstack.org/#/q/project:openstack/neutron+branch:master+topic:bp/address-scopes,n,z 15:43:47 Vikram and Numan have patches up too. 15:44:02 #link https://review.openstack.org/#/c/197552/ 15:44:11 nice! 15:44:56 Let’s get these reviewed. Looks like good work. 15:45:48 #topic IPv6 15:46:12 haleyb: HenryG: john-davidge: others: Anything on IPv6 that we haven’t already discussed? 15:47:02 nothing from me 15:47:19 On the PD front, just to say thanks for the reviews again this week. Freshly rebased patches coming soon 15:48:07 john-davidge: will you ping me when they’re ready? I’d like to give a little more priority to them but I miss a lot in the volume of updates I see. 15:48:33 carl_baldwin: Sure thing 15:48:54 #topic Open Discussion 15:49:23 carl_baldwin: volume of PD patches at the moment mostly due to keeping up with IPAM changes 15:50:25 I have to run, thanks everyone 15:50:27 Just a brief note... 15:51:12 I'm continuing work on 'routed' network as per https://review.openstack.org/#/c/198439/, so would appreciate any further input on that. 15:51:33 Going to try to post complete code for this in the next 2-3 days 15:52:24 15:52:45 neiljerram: Watch for an email from me about the routed segments on the ML. mark mcclain has sent me back to the drawing board a little bit. Considering his feedback. 15:53:15 carl_baldwin: will do. (I heard that feedback too...) 15:53:27 #action carl_baldwin to send mail to ML about routed segments 15:54:03 carl_baldwin: FWIW, personally still interested in pluggable IPAM as a possible ingredient there. 15:55:35 neiljerram: That was essentially the feedback, to make IPAM aware of physical topology. We had considered it briefly before. 15:56:01 There are certain things that the model that I proposed allows that may be more difficult with that model. 15:56:16 Perhaps worth a further look, though. 15:57:33 neiljerram: +1 It is worth a further look. 15:57:43 johnbelamaric: What are you thoughts on this? 15:57:58 carl_baldwin: sorry, let me catch up quickly 15:58:43 johnbelamaric: I thought you had some feedback at the mid-cycle but I don’t remember the specifics. 15:59:51 carl_baldwin: seems like a valid use case. we do now allow the port to be used in IPAM decisions, though it is not in the reference driver. in the boot sequence does nova pass the VM associated with the new port? you could get the host there 16:00:39 hi 16:00:44 carl_baldwin: i recall discussing the model, i think I have to catch up on the commetns on Neil's patch 16:00:46 johnbelamaric: I believe so, at least the host is available as binding:host_id in the mechanism driver code 16:00:51 johnbelamaric: Right now, binding to the host happens after port creation and IPAM. This would have to be changed. 16:01:14 carl_baldwin: i see 16:01:20 carl_baldwin: Ah, that's a big spanner in the works! 16:01:22 are you guys almost done, we have a meeting here now =) 16:01:45 I may be fuzzy on the details. 16:01:52 #endmeeting