09:00:43 #startmeeting Dragonflow 09:00:44 Meeting started Mon Dec 12 09:00:43 2016 UTC and is due to finish in 60 minutes. The chair is oanson. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:00:45 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:00:48 The meeting name has been set to 'dragonflow' 09:00:59 Hello, everybody. Who is here for the Dragonflow meeting? 09:01:00 Good morning 09:01:08 Hi 09:01:10 Hi guys 09:01:11 Hi 09:01:25 hello 09:01:46 #info dimak lihi hujie yuli_s xiaohhui In meeting 09:02:01 #topic Ocata Roadmap 09:02:12 IPv6 - lihi, any news? 09:02:50 I was hopimng to upload the first patch of the neighbor discovery, but I'm having some issues with the fullstack. I hope to solve it today/tomorrow 09:03:57 (I'm talking about the neighbor socialization/advertisement ) 09:04:11 All right. 09:04:24 Looking forwards to it 09:04:33 dimak, any news in the SFC end? 09:04:41 I see you have sent many patches 09:05:00 Not much progress since last week, spent a lot of time working on nb API 09:05:25 I'm still playing around with MPLS/ovs, once I get along with it better I'll upload some SFC app code 09:05:36 Is there any place where you organised the patches - which comes before which? 09:05:48 It would help the review process 09:06:15 Both SFC and refactoring nb api have their own BPs 09:06:49 at the moment SFC depends on the NB api patches 09:07:15 The ones that are tagged [WIP]? 09:07:16 maybe put the number into the patch name and we know who is first and who is second.. 09:07:23 I anticipate the latter will get merged before the former 09:07:51 hujie, I think numbering them is a great idea 09:08:01 :) 09:08:06 Very difficult to tell. If they are not strictly related or conflicting, it is better to have them completely separate, 09:08:12 hujie, +1 09:08:43 I make sure each is on their own topic 09:08:56 and one does depent on the other at the moment 09:09:19 dimak, the topic is overriden by Gerrit to be bp/... 09:09:38 oh, you're right 09:09:56 but each has their own bp so its still good 09:10:07 I'll see how I can make it easier to handle 09:10:54 thats all for me 09:10:55 dimak, maybe there is a commit tag you can use? In mailing lists they use PATCH x/y, but the mailing list structure makes it more organised... 09:11:21 +1 09:11:45 Chassis health/Service health report - I see rajivk isn't here, but there are few comments on his spec 09:11:49 Are there any open issues? 09:12:25 All right. 09:12:26 Do we have a plan to support a branch for dpdk fullstack test? 09:12:35 hujie, that's a good idea 09:13:05 Do you have instructions on how to set it up? I'd also like a local.conf for it 09:13:07 So that we could have tests for userspace and kernal space :) 09:14:03 hujie, could you provide a local.conf, and we'll see how to take it from there? 09:14:05 Not yet, just a suggestion :) 09:14:52 I'm new learner in dpdk, maybe Li Ma is good at it 09:15:19 nick-ma_, are you here? do you have a local.conf for dpdk? 09:15:43 We can open a bp on it, maybe someone will adopt it 09:15:51 yes + 1 09:15:59 +1, but I think a bug is enough 09:16:08 ok 09:16:13 yeah, I wouldn't expect much work of it 09:16:19 +1 for bug 09:16:25 Great! 09:16:35 We don't want to see too many bps 09:16:41 hujie, please open the bug 09:16:48 ok 09:16:59 Tap as a service - The spec looks like it's coming along 09:17:06 I have +1ed it. Waiting for other votes 09:17:32 I'll look at it again today 09:17:33 ok I'll review it 09:17:41 Thanks 09:17:42 me too 09:18:14 sNAT - I have spoken with Igor. He will update the spec to use some existing dragonflow features 09:18:22 such as the information encoded in metadata/registers 09:19:21 LBaaS - I am pushing this feature to Pike. It is not coming along fast enough, and the cycle is short. 09:19:33 I'd rather concentrate efforts elsewhere. 09:19:50 I hope denghui continues to work on it in the background 09:20:33 Anything else for roadmap? 09:21:25 All right. Note we have three pending specs for this cycle: TAPaaS, sNAT, and service health 09:21:46 #topic bugs 09:22:40 I don't think there's anything to report here 09:22:55 Many thanks for nick-ma_ for reporting the gate issue over the weekend, and for xiaohhui's prompt fix 09:22:56 yuyang has committed some dpdk bug 09:23:42 It looks like the solution is given in the bug report, so it should be an easy fix. 09:24:01 He assigned it to himself. So we're waiting for the patch. 09:24:10 ok 09:24:31 Anything else for bugs? 09:24:32 I hope you guys can review this patch, so that I can move on the virtual tunnel feature 09:24:33 https://review.openstack.org/#/c/406531/ 09:24:56 xiaohhui, will do 09:25:27 I'll review it today 09:25:33 Thanks oanson 09:25:37 Thanks lihi 09:25:55 #topic Open Issues 09:26:21 I'd like to raise for a vote making the following gate jobs voting: py34, py35, fullstack 09:26:47 py34 and py35 - there is a move to python 3, and I think it's important to make sure at least the unit test pass and that the code doesn't immediately break 09:27:11 +1 09:27:13 fullstack - despite the issues in the past few weeks, I think it is generally stable, and usually shows a real problem when it fails 09:27:41 +1 09:27:48 fullstack is a de facto gate now 09:28:01 +1 09:28:09 xiaohhui, sorry, what? 09:28:41 I meant the patch will not get +1/2 when full stack fails 09:28:50 so it looks a de facto gate to me 09:28:53 of course :) 09:28:57 I vote to promote it 09:29:16 Great! I'll submit a patch to project-config during the week 09:29:40 like hujie said, adding a gate test for dpdk is also a good idea, and maybe python 3 09:29:46 but one step at a time :) 09:30:03 yes 09:30:31 The floor is for the taking 09:30:40 what about the refactor job? 09:30:47 I'd love to get some reviews for NB api refactor, https://review.openstack.org/#/q/status:open+project:openstack/dragonflow+branch:master+topic:bp/refactor-nb-api 09:31:12 dimak, I will look into it asap 09:31:15 df-local-controller\db-store\nb_api\db_consistent\df_object_reflesher............. 09:31:34 dimak, they are marked WIP. Do you want reviews anyway? 09:31:36 I've seen xiaohhui comments and addressed them 09:31:42 I hope hujie and dimak can work together on refactor, there seems some conflict between your works 09:32:02 Now the conflict is much fewer :) 09:32:15 xiaohhui, +1 hujie :) 09:32:57 oanson, they're wip but I want to hear if the ideas there are sound 09:32:58 dimak, hujie, if you need a common location to collaborate, you can create a clone on github and push commits there 09:33:09 dimak, all right. I'll look at them 09:33:14 you posted some comments there and I think I addressed most if not all 09:33:33 pls help review these two patches: https://review.openstack.org/#/c/408591/ and https://review.openstack.org/#/c/409669/ 09:33:44 xiaohhui, I don't think there is much conflict, it serves to make the dispatch easier :) 09:34:34 I haven't check the latest version, that was my impression from last review. I will update my view on them :) 09:35:11 All right. Is there anything else? 09:36:06 Great. Thanks everyone! 09:36:27 #endmeeting