18:02:29 #startmeeting networking_policy 18:02:30 Meeting started Thu May 1 18:02:29 2014 UTC and is due to finish in 60 minutes. The chair is SumitNaiksatam. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:02:31 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:02:33 The meeting name has been set to 'networking_policy' 18:02:52 hi 18:03:10 Hi all 18:03:11 hi 18:03:33 #info agenda https://wiki.openstack.org/wiki/Meetings/Neutron_Group_Policy 18:04:03 #topic GP bp spec review 18:04:13 #link https://review.openstack.org/#/c/89469/ 18:04:29 i think a lot of comments were addressed over the last week 18:04:58 i think we are making good progress and close to converging 18:05:33 questions/thoughts on that? 18:06:27 ok we can move on the PoC update 18:06:29 looking good! 18:06:36 banix: ok 18:06:43 #topic PoC Status Update 18:07:15 we had some folks huddling together last tuesday and spending a few hours brainstorming on the PoC 18:07:45 i think a lot of issues got sorted out during that discussion 18:07:52 thanks to all those who participated 18:08:43 #info PoC branch is: https://github.com/noironetworks/neutron-group-policy/tree/group-policy-poc 18:09:30 i made some changes to the model based on the feedback on the gerrit review 18:09:44 SumitNaiksatam: such as? 18:09:57 nothing that we did not discuss on tuesday 18:10:06 i am referring to whatever i did before tuesday 18:10:34 currently we still only have EPG, EP, BD and RD represented in the PoC branch as far as the model is concerned 18:10:38 hope to make progress on that 18:10:58 i think rkukura has made good progress on the mapping driver 18:11:03 rkukura: over to you 18:11:04 SumitNaiksatam: Do those resources in the PoC code all match the current spec? 18:11:14 rkukura: i think so 18:11:25 I didn’t make much progress yesteday, but am today 18:11:49 I hope to push an update late today that create the neutron resources when GP resources are created. 18:11:56 rkukura: sweet 18:12:06 i noticed that classifier and action are not in db_group_policy.py 18:12:25 btw, the group policy to classical neutron resources is being captured in this document (for the PoC): #link https://docs.google.com/a/noironetworks.com/document/d/134P7TJdiIfjPWbmstSTY4vp9E6oRYTFs64ON3thFxhI/edit# 18:12:26 I’ll push that to rkukura/mapping, and we can review and decide when to merge to group-policy-poc 18:12:38 sorry, disconnect for a moment 18:12:41 Louis_: not implemented yet 18:13:16 Louis_: we wanted to get an end-to-end flow going with the EP and the EPG 18:13:35 Louis_: also waiting for some of the discussion to settle down on gerrit 18:13:50 but we should be good to implement those now, its next on my list 18:13:59 thx - is the intent to use ml2 driver mechanisms? 18:14:02 I don’t see any reason progress can’t be made on the other parts of the model in the PoC while the mapping of the EP, EPG, BD, and RD are being completed 18:14:33 Louis_: The GP plugin has a driver API similar to ML2’s driver API 18:14:36 rkukura: yes those things are going in parallel 18:15:01 Louis_: And the plan is for the GP plugin to work with the ML2 plugin’s neutron resources when its mapping driver is used 18:15:16 thx 18:15:30 rkukura: my earlier comment was not that the progress is waiting for the mapping to happen, it was more like our focus is on getting on flow going first 18:15:49 on -> one 18:16:03 SumitNaiksatam: agreed 18:16:11 rkukura: anything else? 18:16:19 Not from me 18:16:38 rkukura: thanks, great progress 18:16:43 s3wong_: over to you 18:16:49 for the service redirect part 18:16:56 can you add an example of consumer/provider and role/capab 18:17:14 Louis_: sure 18:17:25 thx 18:17:44 I started looking at it. The idea is to have an UUID passed, and for the PoC, I will assume this is FW, and will then turn on FW on the router between two EPG 18:18:20 s3wong_: ok, i know this is a tough one for you, since the discussion on the services’ side is very much evolving 18:18:33 s3wong_: i guess we will need to iterate through this 18:18:57 in the future, of course, the service object would tell us where to put the object can be inserted and traffic could be steered 18:19:18 but for now, we have no way of "redirect" flow to "service" in Neutron 18:19:31 will the redirect action value be a list of service instance uuids or a service chain uuid or either? 18:19:41 so for the PoC, I will just turn on FWaaS on a router :-) 18:19:49 s3wong: yes, we will rely on abstractions exposed by the adv services’ framework to do insertion, steering, etc 18:19:54 s3wong: ok 18:20:05 Louis_: a service chain uuid 18:20:09 Louis_: for now, we agreed on having redirect be the final action 18:20:10 Louis_: either one service or a service chain 18:20:12 Louis_: not implemented yet 18:20:26 and that object can certainly be service chain 18:20:59 Louis_: yeah, like banix and s3wong said abouve, its one single uuid, either service or service chain 18:21:01 what do u mean by "final" action? 18:21:14 Louis_: however service chain is currently not represented in neutron 18:21:26 i see 18:22:59 any other questions for s3wong or on service redirect? 18:23:08 redirect action that is 18:23:10 would order of the list of service uuids be siginifcant? 18:23:33 Louis_: the order is significant in the proposed chain resources 18:23:36 *resource 18:23:54 ideally it would be a graph representation 18:24:02 but that is difficult to achieve 18:24:15 SumitNaiksatam: Louis_: are we still talking about 'redirect' or service chaining in Neutron in general? 18:24:18 we will be happy if we can get going with a linear ordered chain 18:24:37 s3wong: yes, sorry, we are digressing to advanced services discussion in neutron 18:24:40 redirect 18:25:05 Louis_: 'redirect' to only one UUID, and you should only have one 'redirect' action 18:25:17 at least for now 18:25:25 ok 18:25:36 Louis_: fyi - we have a different meeting to discuss services chains and other services’ related aspects in neutron: #link https://wiki.openstack.org/wiki/Meetings/AdvancedServices 18:25:54 am aware of that thx 18:26:33 hemanthravi: any progress on the Client, CLI? 18:27:14 SumitNaiksatam: not much since the last commit, will work on this and should have it done by sun 18:27:27 hemanthravi: ok 18:27:40 hemanthravi: so what resources are currently implemented? 18:27:51 and test it out with rkukura's mapping branch, any config that i need to have? 18:28:07 endpoint, endpoint-group are currently impl 18:28:13 hemanthravi: we have a PoC branch 18:28:30 hemanthravi: are they updated with the latest model? 18:29:00 SumitNaiksatam: i think ep, epg didn't change, but will check 18:29:15 hemanthravi: ok thanks 18:29:28 rkukura: any config that i need to run the gp plugin? 18:29:43 hemanthravi: any chance that you can have the EP/EPG by friday? 18:29:47 I haven’t tried it outside the UTs 18:30:07 hemanthravi: the service plugin needs to be set to the group policy plugin 18:30:09 SumitNaiksatam: most likely sat 18:30:19 You’ll need to configure a core_plugin and list of service_plugins, which should include group_policy plugin 18:30:36 rkukura, SumitNaiksatam: ok 18:31:05 hemanthravi: there is a noop plugin driver which is used for the UTs 18:31:12 You may need to configure the GP plugin to use the dummy driver for now, or the mapping driver if you want neutron resources to be created (once I push) 18:31:36 hemanthravi: yeah rkukura and I are saying the same thing 18:32:00 rkukura: ok will try it out and ping you with any qs 18:32:00 hemanthravi: i would recommend that you just use the dummy driver to begin with 18:32:09 SumitNaiksatam: will do 18:32:19 yes, the noop driver’s name is ‘dummy’ 18:32:35 hemanthravi: sure 18:32:40 any questions for hemanthravi regarding client/CLI 18:33:50 ok moving on 18:34:00 ronak here? 18:34:27 so ronak bravely stood up to take on the Horizon piece! 18:34:39 * SumitNaiksatam applauds Ronak :-) 18:34:49 +1 18:34:52 but he doesnt seem to be here to absorb the applause 18:34:53 * mandeep agrees 18:35:30 hi I'm writing wireframes 18:35:31 for horizon 18:35:36 https://docs.google.com/presentation/d/1SmbhY5GTBKFV0U6XmAlaWn2nm-biV5bFVZDURZslNrg/edit#slide=id.g333b16535_0351 18:35:45 nati_ueno: oh wow! 18:35:54 * s3wong applauds nati_ueno instead :-) 18:36:03 ronak also had captured some of the discussions in his wire frames 18:36:11 I'm still struggling to understand gp, but we can use this as basement for discussion 18:36:28 so can I have feedback on here? 18:36:29 nati_ueno: roank is also building wireframes for group based policy. Can you follow up with him? 18:36:40 nati_ueno: this is very nice and detailed 18:36:43 mandeep: yap, interstting 18:36:51 is here here? 18:37:13 mandeep: could you share his mail address? 18:37:20 SumitNaiksatam: thanks 18:37:23 nati_ueno: good work on the UI Prototype 18:37:35 nati_ueno: I will send that tou you by email 18:37:42 mandeep: Thnaks 18:37:45 Swami_: Thanks 18:37:48 so quick question 18:37:52 what's filters? 18:37:59 I can't figureout, so it is ? in the wireframe 18:38:14 nati_ueno: thanks for the work; had a quick look; one comment 18:38:24 nati_ueno: filters defines who can consume a part of a contract, and who can provide a part of a contract 18:38:29 nati_ueno: filters limit the scope of a specific policy rule to a specific role/capability (if that us desired) 18:38:47 nati_ueno: but filters can be optional 18:39:08 nati_ueno: ahhh i noticed teh answer to the question i was about to ask…. 18:39:26 hmm filters 18:39:31 banix: ok 18:39:35 nati_ueno: quickly skimming through the document, i think we can simplify some of the workflow 18:39:54 SumitNaiksatam: ok 18:40:09 SumitNaiksatam: also, could you update filters part? 18:40:22 SumitNaiksatam: may be, I can understand it if I see the UI 18:40:32 nati_ueno: sure 18:40:49 SumitNaiksatam: I added you as an editor 18:40:55 nati_ueno: thanks 18:41:31 nati_ueno: Can you add me and ronal as editors as well 18:41:32 SumitNaiksatam: now you should believe me as I'm app guy.. actually, I'm UI guy :P 18:41:58 nati_ueno: Renaissance man ... ;-) 18:41:58 mandeep: sure! please share mail address of yours (gmail version) 18:42:00 nati_ueno: you are all rolled into one 18:42:12 volunteering my services... 18:42:14 nati_ueno: you talented Mr. Rippley! 18:42:21 nati_ueno: I will send you that in the email as well 18:42:32 SumitNaiksatam: mandeep: Thanks 18:43:28 nati_ueno: now that you mentioned you are an app guy, what do you think of terms “provider” and “consumer”? 18:43:52 nati_ueno: do they sound right? Do people use other terms? 18:43:59 banix: honestly, it takes time to figure out the meanings 18:44:35 we need some analogies 18:44:41 any more questions for on Horizons 18:44:44 nati_ueno: good point 18:44:45 need some examples 18:45:20 Louis_: There is a PoC use-case on the wiki. Did you have a chance to look at that? 18:45:44 yes could be clearer 18:46:01 so my big concern is there is no resource which shows consumer - provider -contract reationship. 18:46:13 it is devised in several resource 18:46:32 In UI, we can correlate it, but 18:46:38 it sounds like hard in CLI 18:46:49 nati_ueno: actually there is are two resources to capture the relationship 18:46:56 SumitNaiksatam: yes two 18:47:02 SumitNaiksatam: so it is devided 18:47:13 SumitNaiksatam: ConsumerScore, ProviderScope, right? 18:47:13 nati_ueno: no, one for provider, one for consumer 18:47:22 nati_ueno: yes 18:47:32 SumitNaiksatam: and ConsumerScope and ProviderScope is linked by selector? 18:47:36 nati_ueno: we got a lot of feedback that it was easier to understand that way, then all rolled into one 18:47:59 ya, 18:48:11 so if we have one single scope object 18:48:14 nati_ueno: they both have a selector as an attribute 18:48:26 SumitNaiksatam: yes. so they are connected by attribute 18:48:27 nati_ueno: i originally had only one 18:48:38 nati_ueno: One important requirement was that these actions be decoupled from each other in time and privilages. This separation enables that 18:49:06 mandeep: we can have multiple scopes 18:49:50 one thing i wanted to put out there is that the workflow and interface we see in the UI does not have to map the exact resources in the backend 18:49:52 nati_ueno: I am sorry I did not follow that. 18:50:08 we can devise the workflow in an easier to consume fashion 18:50:14 nati_ueno: Scope is a property of the relationship (either provider or consumer), 18:50:14 SumitNaiksatam: It is better to easy to map 18:50:39 Good model is easy to map UI 18:51:09 nati_ueno: not necessarily 18:51:14 we have 10 mins or less 18:51:29 Is rudra here 18:51:51 thanks nati_ueno for the update 18:51:57 moving on to heat 18:51:58 SumitNaiksatam: Thanks! 18:52:00 prasadv: here? 18:52:14 i believe prasadv claimed victory here :-P 18:52:17 nati_ueno: thanks! 18:52:21 yes 18:52:32 I have the classifier template 18:52:50 and rest of them are mostly copy and paste 18:52:56 banix: :) 18:53:03 I had a question as to testing these 18:53:18 prasadv: go ahead 18:53:19 when do we want to test the first call 18:53:34 prasadv: whenever the client is ready 18:53:41 i mean say classifier through the neutron api 18:53:49 prasadv: that will be sometime next week 18:53:52 prasadv: Rudra is bringing up an AWS instance with devstack on the integration branch 18:53:53 this is not with the client but with neutron api 18:54:10 prasadv: We will be doing ene-to-end PoC integration on that 18:54:41 ok. meanwhile I will check in the resources.. 18:54:50 prasadv: please do 18:55:00 prasadv: cool 18:55:07 prasadv: thanks, and great progress! 18:55:17 also the template I will be using is what is provided on the preso right? 18:55:41 will there be a gp meeting at summit? 18:55:50 Louis_: good question 18:55:55 that was the next topic 18:56:01 thanks prasadv ! 18:56:03 Louis_: both conference presentation as well as design summit, yes 18:56:17 Yes, thanks prasadv 18:56:22 date time? 18:56:25 #topic Atlanta summit 18:56:56 design summit session #link http://junodesignsummit.sched.org/event/e9dd467daf0cae0cdb29ee97d64bbf56#.U2KMDK1dX3A 18:57:22 Louis_: presentation will be Thurs @13:30; design session keeps on changing :-) 18:57:32 Thursday at 1:30 for the general talk 18:57:35 conference session: #link http://openstacksummitmay2014atlanta.sched.org/event/456a216328d5ff2162e3cefa168eb648#.U2KZE61dX3B 18:58:01 we need to prepare for both 18:58:15 banix: over to you for the preso 18:58:42 have a first draft for the conference talk 18:59:09 banix: nice 18:59:18 i think its coming along very well 18:59:19 Trying to keep the discussion as simple as possible and then build on it to avoid getting bugged down in details early on 18:59:19 banix: Thanks! 18:59:45 banix: thanks! 18:59:53 #topic Open Discussion 18:59:57 any parting thoughts? 18:59:58 Sure; Will be working on it in the next 2 weeks :) 19:00:11 banix: thanks 19:00:19 banix: up until the morning of the presentation! 19:00:26 s3wong: :-) 19:00:28 s3wong: indeed :) 19:00:31 banix: it is exactly two weeks! 19:00:44 s3wong: thats scary! 19:00:57 banix: s3wong: Just wanted to mention that we should have a common devstack server to work on soon 19:00:57 anything else? 19:01:10 mandeep: thanks for bringing that up 19:01:15 rudra, is not here 19:01:19 mandeep: that would be great 19:01:22 he has taken that up 19:01:23 mandeep: great! 19:01:31 * SumitNaiksatam applauds rudra! 19:01:43 ok we over time 19:01:46 thanks everyone! 19:01:51 until next week 19:01:53 bye! 19:01:54 thanks! 19:01:55 bye 19:01:55 #endmeeting