15:02:00 #startmeeting neutron_l3 15:02:01 Meeting started Thu Aug 21 15:02:00 2014 UTC and is due to finish in 60 minutes. The chair is carl_baldwin. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:02:02 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:02:05 The meeting name has been set to 'neutron_l3' 15:02:10 carl_baldwin: Hi 15:02:16 #topic Announcements 15:02:23 #link https://wiki.openstack.org/wiki/Meetings/Neutron-L3-Subteam 15:02:33 Feature Proposal Freeze is August 21st (today) 15:02:50 Patches for new features posted after today will be automatically deferred. 15:03:34 Juno-3 is September 4th. That is just two short weeks away. 15:04:04 Reviewers will be getting in to even more of a crunch as that approaches. 15:04:12 #link https://wiki.openstack.org/wiki/Juno_Release_Schedule 15:04:56 #topic Bugs 15:05:31 I think the bugs have been kept under control. There are a couple I’d like to get in to Juno. 15:05:40 bug 1334926 15:05:53 * pcm_ sorry late 15:05:55 bug #1335375 15:06:11 #link https://bugs.launchpad.net/neutron/+bug/1334926 15:06:17 #link https://bugs.launchpad.net/neutron/+bug/1335375 15:06:35 I think that is all for the general bugs. 15:06:52 #topic neutron-ovs-dvr 15:07:19 Swami: ping 15:07:37 I don’t see Swami around. Does anyone else have a report for DVR? 15:07:59 Mostly working through bugs 15:08:41 I made good progress on migration yesterday 15:08:49 I should post an update to the patch today 15:08:59 Initiated design discussions with FWaaS team 15:09:06 there are dependencies with the SNAT bugs however 15:09:36 I merged the existing SNAT fix patch into my local migration patch 15:09:56 mrsmith: merged? Or did you rebase yours to it? 15:10:10 hand merged 15:10:17 as a prototype 15:10:38 so we may want to consider making the migration patch dependent in gerrit for the snat fixes 15:10:53 or just make sure the snat patch merges soon 15:11:02 mrsmith: That will be better in the long run. 15:11:49 I have updates to the snat patch also 15:11:58 better handling for subnets 15:12:05 in the 'do vms exist' method 15:12:06 We have 19 bugs listed in the backlog. Most are medium importance so I’m not too concerned about that. But, I find it harder to keep up with the reviews because it is hard to tell when is ready for review. 15:12:29 true 15:12:56 in my opinion the snat and migration patches weren't ready *yet* but I think they will be on the next revs 15:13:14 they both need more UTs also 15:14:10 mrsmith: okay. 15:15:31 mrsmith: Swami: Rajeev: viveknarasimhan: Please don’t hesitate to ping me when you feel a patch is ready for review. 15:15:55 thanks. sounds good. 15:16:09 There is one High Importance bug that does not have an owner. 15:16:26 carl_baldwin: Can you share your search query? 15:16:36 #link https://bugs.launchpad.net/neutron/+bug/1358554 15:16:44 amuller: The query for bugs? 15:16:49 yeah 15:17:11 I have two that are linked from here: https://wiki.openstack.org/wiki/Neutron/DVR/HowTo#DVR_Backlog 15:17:23 thanks 15:17:56 amuller: glad to help. 15:19:16 I will find someone to drive bug 1358554 15:20:09 #action carl_baldwin will find owner for bug 1358554 15:20:13 carl_baldwin: I can give it a try 15:20:14 Anything else on DVR? 15:20:35 carl_baldwin: will probably need some ramp up time, don't know the code very well 15:20:57 for 1358718 15:20:59 rossella_s: Are you familiar with how the dvr experimental job has evolved since July? 15:22:11 carl_baldwin: to be honest not very much...but I always wanted to catch up and I have some time now...anyway if somebody with more experience is willing to take it, please go on 15:22:51 rossella_s: I can catch you up a little after the meeting since we’ll be chatting anyway. 15:23:02 * carl_baldwin goes to look at 15:23:05 carl_baldwin: great, thanks! 15:23:09 * carl_baldwin 1358718 15:23:54 Rajeev: ? 15:24:10 https://bugs.launchpad.net/neutron/+bug/1358718 15:24:39 is this required for juno-3 ? 15:25:41 Rajeev: It is marked Medium so technically no. The Highs should be worked down first. 15:25:53 agreed. 15:26:18 hi 15:26:25 sorry I am late. 15:26:27 Swami: hi 15:27:32 With that said, I’d like very much to wittle down the DVR backlog. The number of Mediums is growing steadily. 15:27:39 Swami: Anything to report? 15:27:48 nothing to add at this time. 15:28:09 carl_baldwin: can we chat later on https://bugs.launchpad.net/neutron/+bug/1335375 you mentioned earlier 15:28:22 carl_baldwin: Swami: I'd love to get a decision down on the CLI patches for DVR 15:28:31 Sort it out... We can talk about it after the meeting 15:28:57 amuller: sure, 15:29:51 Rajeev: yes, we can chat. Finding me a bit later today. 15:30:39 Anything else to discuss? 15:31:06 #topic l3-high-availability 15:31:12 amuller: safchain: hi 15:31:35 Sylvain is on vacation 15:31:46 I brushed up the server side patches since last week 15:32:18 The scheduler patch is the weakest in the chain... There's a DB query there that's untested and I wouldn't be surprised if it's wrong 15:32:24 The rest of the patches are better 15:32:35 amuller: I think I saw that DB query. 15:32:37 ;) 15:32:39 We need reviews, I'll iterate fast and hopefully we'll be able to merge the feature 15:33:03 I pushed a bunch of patches for another feature to observe the freeze today, I'm back on L3 HA now 15:34:46 I’ll watch for you to address feedback. 15:34:51 I will 15:34:55 amuller: Do we need to chat about the DVR integration? 15:34:59 You might force me to work on Friday =D 15:35:21 carl_baldwin: I don't have any knowledge yet to contribute on that front sadly 15:35:26 But I will by early next week 15:35:40 I see two issues (there may be more) 15:36:03 First is that the scheduler seems to be written for one or the other and not both. 15:36:11 right 15:36:35 Second may be a datapath issue since the snat namespace uses a second port that is not the gateway port. 15:37:28 I'll have to look in to that 15:38:01 So, feature freeze is two weeks out which concerns me a lot at this point. 15:38:13 Yeah I have a bunch of new grey hair now 15:38:53 carl_baldwin: the second port for snat - do you mean the internal port? 15:39:02 mrsmith: yes. 15:39:08 k 15:39:35 vrrp would have to work on that port and not the internal gateway port. 15:39:48 I shouldn’t have called it a gateway port. ;) 15:40:11 I mean the internal port that is the gateway for internal traffic. 15:40:48 amuller: Let’s talk a little more about this soon when you’ve had a chance to think it over. 15:40:54 Will do 15:41:05 Meanwhile, I’ll move on the the agent side reviews. 15:41:45 The l3 functional test is failing because of a gate configuration issue, there's patches by Maru to fix that 15:41:49 it works locally 15:42:31 And the agent-side patch that adds HA capabilities adds a new functional test for HA routers, which suffers from the same issue 15:42:51 amuller: do you have links to those patches by marun ? 15:43:17 in Neutron: https://review.openstack.org/#/c/114717/ 15:43:27 in infra: https://review.openstack.org/#/c/114416/ 15:44:12 Thanks, I will review. 15:44:18 amuller: Anything else to discuss? 15:44:24 Nay 15:44:31 Thanks for your time Carl 15:45:25 amuller: thank you. 15:45:38 #topic bgp-dynamic-routing 15:45:44 devvesa: Hi 15:45:47 hi 15:46:00 I see you’ve been posting code. 15:46:22 I plan to start reviewing it today. 15:46:23 yes, I've splitted the code in three patches 15:46:27 great! 15:46:45 i've edited the wiki subteam page with the links to the reviews 15:47:02 How much have you been able to get working in your testing? 15:47:16 devvesa: Thanks for the links. I’ll go there to start. 15:47:41 extension and database mixin is fully tested 15:48:04 Great. 15:48:18 i've been testing the agent with a remote quagga exchanging routes , and it works 15:48:41 Did the bgp speakers comparison discussion conclude? I’m sorry I’ve been out in DVR/HA land for a long time. 15:48:41 i need to spend more time testing more complex scenarios like two agents working in HA 15:49:05 well, I've just used Ryu because the api is clean and it has basic features 15:49:30 #link https://wiki.openstack.org/wiki/Neutron/BGPSpeakersComparison 15:49:36 Ryu does look like a stand-out. 15:49:46 How is the ryu code delivered? 15:50:00 installing from pypi 15:50:13 Okay, that should be good. 15:50:31 and i like how it exchanges routes without touching the routing table of the agent's machine 15:50:47 devvesa: sounds like you’ve done a lot. 15:51:03 devvesa: I agree. When I was playing around with quagga, that was a pain. 15:51:13 This sounds like just what we need. 15:51:19 devvesa: I think a 'how to test' wiki page but would really helpful 15:51:38 amuller: ok! I'll do 15:51:50 amuller: +1 15:52:05 devvesa: great. 15:52:06 #action devvesa write a wiki page about dynamic routing and test 15:52:36 amuller: don't expect a blog post like yours :) 15:53:14 I'd appreciate something shorter hehe 15:53:32 #action carl_baldwin will review BGP patches 15:53:34 don't worry about this! :D 15:53:50 devvesa: anything more? 15:54:22 just a couple of comments, I don't know if you will be agree: 15:54:50 1. as I told you, I've tried to simplify the spec doing less exposed endpoints 15:55:14 if you are not agree I have the 'extended version' :) 15:55:19 Yes, I think #1 should be fine as we discussed. 15:55:46 and 2. the discovered routes are applied over router's static routes 15:56:22 but static routes are applied as a whole, so once you associate a router to a routing instance, it will override any routes 15:56:30 you have previously 15:56:47 I don't know if we can live with this 15:57:13 So, using extraroutes and dynamic routes is mutually exclusive? 15:57:25 yes 15:57:54 so you can modify extra routes from the API, but the periodic task that check discovered routes will override it 15:58:13 only in the router you have associated 15:58:22 for dynamic routing 15:58:25 I see. 15:59:21 maybe adding a new attribute in the extraroutes entity with the source that added it (static or dynamic) 15:59:24 I might have to think about that. My initial reaction is that it may be acceptible. However, I wonder if some API tweak to detect and raise an error in the event that both are applied to a router. 15:59:24 would be enough 15:59:43 but I have not done this part 16:00:18 Is this a problem if dynamic routing is turned on but a router does not learn routes. Only advertises routes? 16:00:37 you can do it, sure 16:00:52 devvesa: Let me think on it while I review the code. 16:00:54 We’re out of time. 16:01:02 ok, thanks carl 16:01:11 Bye everyone. Thanks for all your work. 16:01:14 #endmeeting