03:00:06 <hongbin> #startmeeting zun 03:00:07 <openstack> Meeting started Tue Dec 13 03:00:06 2016 UTC and is due to finish in 60 minutes. The chair is hongbin. Information about MeetBot at http://wiki.debian.org/MeetBot. 03:00:08 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 03:00:10 <openstack> The meeting name has been set to 'zun' 03:00:12 <hongbin> #link https://wiki.openstack.org/wiki/Zun#Agenda_for_2016-12-13_0300_UTC Today's agenda 03:00:17 <hongbin> #topic Roll Call 03:00:21 <Namrata> hi 03:00:21 <pksingh> pradeep 03:00:24 <kevinz> kevinz 03:00:34 <Namrata> Namrata 03:00:38 <eliqiao> hi 03:00:41 <eliqiao> eliqiao 03:00:44 <mkrai_> Madhuri Kumari 03:00:54 <shubhams> shubhams: 03:00:58 <Wenzhi> Wenzhi 03:01:09 <hongbin> thanks for joining the meeting Namrata pksingh kevinz Namrata eliqiao mkrai_ shubhams Wenzhi 03:01:19 <hongbin> #topic Announcements 03:01:24 <hongbin> 1. Welcome Pradeep to the core team 03:01:31 <Wenzhi> welcome! 03:01:39 <mkrai_> Congratulations pksingh! 03:01:43 <shubhams> Congrats Pradeep 03:01:45 <eliqiao> great addition! 03:01:45 <hongbin> pksingh: thanks for your contribution. it is good to have you in the core team 03:01:47 <mkrai_> Welcome 03:01:58 <pksingh> thanks all :) 03:02:06 <hongbin> #topic Review Action Items 03:02:12 <hongbin> 1. hongbin create a etherpad to discuss the zun core api (DONE) 03:02:18 <hongbin> #link https://etherpad.openstack.org/p/zun-core-api 03:02:29 <hongbin> we will revisit this etherpad later in the agenda 03:02:35 <hongbin> 2. hongbin start a ML to discuss the k8s integration bp (DONE) 03:02:42 <hongbin> #link http://lists.openstack.org/pipermail/openstack-dev/2016-December/108569.html 03:03:02 <hongbin> shubhams replied to the ML 03:03:11 <hongbin> so, let's continue the discussion here 03:03:28 <hongbin> #topic Kubernetes integration (shubhams) 03:03:36 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/k8s-integration The BP 03:03:41 <hongbin> #link https://etherpad.openstack.org/p/zun-k8s-integration The etherpad 03:04:03 <hongbin> shubhams: want to drive this one? 03:04:14 <shubhams> hongbin: Had discussion with mkrai_ and pksingh and we think that having a pod (or similar concept) within zun is better than just acting like a proxy 03:04:47 <hongbin> shubhams: ack 03:04:51 <shubhams> We want to know views of team as it will require considerable efforts and analysis if we do it 03:05:22 <hongbin> thoughts everyone? 03:05:25 <diga> o/ 03:05:37 <hongbin> welcome to the meeting diga 03:05:48 <diga> hongbin: thnks :) 03:05:55 <Wenzhi> acting like a proxy is definitely not a good idea 03:06:04 <mkrai_> +1 for it as I think behaving as proxy to any COE wouldn't add much value to zun 03:06:19 <eliqiao> +1 for not using proxy as k8s 03:06:44 <hongbin> i also agree on this: no proxy 03:06:46 <pksingh> and having purpose of comon interface for different coe will be defeated if we act like proxy 03:07:35 <hongbin> i think some team members also challenged the necessary to bring k8s to zun 03:07:35 <shubhams> #link http://lists.openstack.org/pipermail/openstack-dev/2016-December/108735.html 03:07:44 <sudipto> o/ sorry for being late. 03:07:45 <mkrai_> +1 pksingh 03:07:50 <hongbin> sudipto: hey 03:07:51 <sudipto> logged in from the car :) 03:08:01 <hongbin> sudipto: :) 03:08:12 <hongbin> sudipto: we just discussed the k8s integration bp 03:08:14 <shubhams> Link I posted has the reasoning for our suggestion. Interested people can see :) 03:08:27 <hongbin> sudipto: do you have any comment about that? 03:09:20 <hongbin> shubhams: i think we can break this bp into two : 1. introduce pod, 2. k8s integration 03:09:32 <shubhams> hongbin: agree 03:09:38 <pksingh> hongbin: +1 03:09:39 <hongbin> shubhams: but we can start them in parallel 03:09:49 <shubhams> hongbin: And I think both should work in parallel 03:09:51 <hongbin> #action hongbin split the k8s bp into two 03:10:20 <Wenzhi> if we introduce pod, do we need to introduce k8s's RC and service as well? 03:10:44 <hongbin> Wenzhi: i don't think we should do it right now 03:10:55 <shubhams> Wenzhi: ATM, I dont think we should focus on having them 03:11:02 <Wenzhi> ok 03:11:20 <hongbin> the last question i have for all of you 03:11:49 <hongbin> do you think if it is valuable to bring k8s to zun (since several poeple challenged this idea in before) 03:12:06 <hongbin> just want to make sure if everyone think this is the right direction 03:12:33 <hongbin> any opposition point of view? 03:12:56 <sudipto_> I have POVs, but i am not hard bound. 03:13:17 <mkrai_> As of now I would say "no". When we have our own pod like concept we can eventually integrate k8s also without affecting the end users. 03:13:25 <hongbin> sudipto_: yes, we would like to hear your point of view 03:13:36 <sudipto_> I feel that k8s integration is as good as asking people to use another abstraction on top of kubernetes and us chasing the kubernetes releases to have feature parity. 03:13:46 <mkrai_> All COEs will act yes the drivers and we can plug them in as per use. 03:14:03 <shubhams> I think having k8s is just optional , and not necessary for zun's success . 03:14:29 <sudipto_> yeah and i think the trade off is, whether we should work on optional features or the relatively important ones. 03:14:51 <sudipto_> but if there's a real need for k8s, i would like to hear that. 03:15:48 <diga> hongbin: mkrai_ : Ultimate target is to use zun to provision/manage containers on top of Container infrastructure like k8s, docker swarm etc 03:16:31 <hongbin> ok 03:16:38 <shubhams> hongbin: IIRC, when you came back from Barcelona Summit, you said that people are interested in k8s with zun. What was their motivation for this(if by any chance you could discuss in details) 03:17:07 <hongbin> shubhams: they just kept asking if zun support k8s 03:17:22 <mkrai_> diga: Initially we planned to integrate with k8s, swarm but now it seems people are not in favor of it 03:17:23 <shubhams> hongbin: ok 03:18:09 <hongbin> shubhams: the one that really asked for k8s is other openstack team (i.e. trove). they wanted a unified api for different coe 03:18:19 <sudipto_> My take is, i am not opposed to supporting k8s but it's good to build on the docker driver to support composition of containers first. 03:18:33 <diga> mkrai_: okay 03:19:00 <mkrai_> +1 sudipto_ 03:19:02 <Wenzhi> sudipto_: agreed 03:19:03 <sudipto_> else we could have just stuck to being a gateway for the COEs 03:19:04 <hongbin> ok, sounds like we should adjust the priority of the k8s bp? 03:19:15 <shubhams> hongbin: agree 03:19:20 <kevinz> +1 sudipto_ 03:19:33 <hongbin> right now, the bp is essential, let's drop it to medium? 03:19:50 <mkrai_> +1 03:19:56 <diga> +1 03:20:01 <hongbin> done 03:20:21 <hongbin> i will create a pod bp, which will be given a higher priority 03:20:26 <sudipto_> +1 03:20:35 <hongbin> sounds good to everyone? 03:20:40 <Wenzhi> good 03:20:44 <hongbin> any other comment? 03:20:46 <mkrai_> Yes 03:20:56 <Qiming> +3 03:21:01 <diga> fine 03:21:02 <hongbin> Qiming: :) 03:21:03 <shubhams> hongbin: sounds good 03:21:07 <hongbin> very good 03:21:10 <pksingh> +1 03:21:25 <hongbin> next one 03:21:26 <hongbin> #topic Support interactive mode (kevinz) 03:21:32 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/support-interactive-mode The BP 03:21:38 <hongbin> #link https://review.openstack.org/#/c/396841/ The design spec 03:21:41 <hongbin> kevinz: ^^ 03:21:50 <kevinz> Hi 03:23:02 <hongbin> kevinz: do you like to give an update for this one? 03:23:03 <kevinz> This week I'm still working to give protocol,,,so I don't have much more update this week. Hope to submit first draft before next meeting 03:23:17 <hongbin> kevinz: ack. that is fine 03:23:22 <kevinz> Thanks~ 03:23:33 <hongbin> next topic 03:23:50 <hongbin> #topic Should we hold a team meeting at 2016-12-27 UTC 0300? 03:24:40 <mkrai_> hongbin: Why we don't want to hold meeting? 03:24:45 <shubhams> I will be available 03:24:55 <hongbin> it is xmas 03:24:55 <pksingh> I am available too 03:25:03 <shubhams> mkrai_: many people will be gone for new year leave (i guess) 03:25:09 <sudipto> i will be off. 03:25:09 <mkrai_> I will be available :) 03:25:19 <kevinz> I Will be off too 03:25:20 <Wenzhi> available 03:25:29 <sudipto> Going to thailand :) 03:25:41 <pksingh> sudipto: :) 03:25:51 <hongbin> ok, sounds like half of hte team members will not be available 03:26:03 <hongbin> let me ask another question 03:26:09 <hongbin> #topic Should we hold a team meeting at 2017-01-03 UTC 0300? 03:26:27 <mkrai_> available 03:26:28 <hongbin> will everyone available at Jan 03? 03:26:28 <sudipto> Available for that one :) 03:26:33 <kevinz> Fine 03:26:35 <shubhams> available 03:26:36 <pksingh> available 03:26:43 <diga> yep 03:26:49 <Namrata> available 03:27:02 <hongbin> ok, then how about dropping the 12-27 one, then hold the 01-03 03:27:18 <mkrai_> Ok 03:27:21 <sudipto> would you be available hongbin 03:27:21 <pksingh> k 03:27:21 <sudipto> ? 03:27:40 <hongbin> sudipto: i am not sure, it is a holiday at Canada :) 03:27:59 <hongbin> sudipto: but Jan 03, i will be available 03:28:20 <sudipto> hopefully not logging in from a motor vehicle like me right now :) 03:28:27 <hongbin> #agreed cancel the team meeting at 12-27 03:28:35 <hongbin> sudipto: :) 03:28:47 <hongbin> ok 03:28:51 <hongbin> #topic Open Discussion 03:29:13 <hongbin> there is an etherpad that needs to be discussed: #link https://etherpad.openstack.org/p/zun-core-api 03:29:15 <hongbin> #link https://etherpad.openstack.org/p/zun-core-api 03:29:38 <diga> hongbin: I would like to take this BP - https://blueprints.launchpad.net/zun/+spec/coe-integration 03:29:40 <hongbin> we can do it right now, or leave it as homework 03:30:00 <hongbin> diga: this bp is of priority "not" 03:30:05 <diga> We need to generic layer may be based on the factory implementation 03:30:16 <Qiming> just posted some comments to the core api 03:30:20 <hongbin> diga: which means it is not in the recent roadmap, it is a long term idea 03:30:30 <diga> hongbin: okay 03:30:56 <hongbin> diga: however, you can work with mkrai_ shubhams pksingh for the k8s bp, it is one of the coe 03:31:07 <diga> hongbin: anything I can start 03:31:10 <diga> hongbin: okay 03:31:34 <hongbin> Qiming: you are of color blue? 03:31:47 <Qiming> I cannot read colors, :) 03:31:54 <Qiming> yes, probably 03:32:53 <Qiming> my suggestion is to merge all container operations into a single URL 03:32:58 <hongbin> Qiming: you were suggestion: /containers/<ID>/actions/<ACTION> ? 03:33:10 <Qiming> yes 03:33:19 <hongbin> Qiming: ack 03:33:26 <Qiming> http://developer.openstack.org/api-ref/compute/#servers-run-an-action-servers-action 03:34:14 <hongbin> others, thoughts? 03:35:04 <hongbin> Qiming: i see. nova is designed the api in this way 03:35:13 <pksingh> hongbin: i think action is part of the body? 03:35:16 <Qiming> also, for these operations, maybe we should return 202 instead of 200 03:35:30 <hongbin> pksingh: yes, <action> is in the body 03:35:57 <hongbin> that means make them async? 03:36:11 <Qiming> yes, they are inherently async operations 03:36:37 <hongbin> i see 03:36:43 <mkrai_> Qiming: the benefit is reduced no of APIs exposed? 03:36:54 <Qiming> yes, mkrai_ 03:36:57 <mkrai_> Single API endpoint for all actions 03:37:13 <Qiming> except for basic CRUD calls 03:37:22 <mkrai_> Yes got it 03:37:36 <Qiming> a POST to the collection 'actions' means a creation of an action resource 03:37:38 <pksingh> Qiming: isn't it confusing having same url for different controller actions? 03:38:01 <Qiming> that is a more RESTful way to operate your resources 03:38:23 <Wenzhi> right 03:38:40 <Qiming> it is different from CLI or GUI 03:39:04 <Qiming> from CLI or dashboard, you can still present these operations in whatever way users appreciate 03:39:46 <Qiming> just my 2 cents for team to consider 03:39:51 <csomerville> It might be worth asking the nova team if they're happy with that design decision. 03:40:00 <eliqiao> suggest to fellow API design as https://wiki.openstack.org/wiki/API_Working_Group 03:40:44 <eliqiao> s/fellow/follow/ 03:41:07 <pksingh> csomerville: +1 03:41:26 <hongbin> ok, i can create a bp for that first 03:41:30 <Qiming> there was a patch #234994 proposing this idea 03:42:02 <Qiming> however, it was stuck by some trivial things, and blocked by the tendency people loves "-1" a patch 03:42:07 <csomerville> Was just discussing today how multiple actions to single URL in nova makes it more difficult to track feature usage by http access logs 03:42:11 <hongbin> then, everyone can cast feedback to the bp (i.e. what is nova POV) 03:42:41 <Qiming> csomerville, reasonable concern though 03:42:54 <mkrai_> hongbin: Qiming How about raising ML with nova in subject? 03:43:08 <Qiming> sounds tood 03:43:09 <pksingh> mkrai_: +1 03:43:10 <Qiming> good 03:43:15 <hongbin> ok 03:43:31 <hongbin> mkrai_: maybe raise it with API WG is more appropreciate 03:43:38 <mkrai_> Yes 03:44:05 <hongbin> #action hongbin raise a ML with API WG to discuss the consolidate of actions into single URL 03:44:33 <hongbin> at the same time, we can ping the nova channel to get more feedback 03:45:01 <hongbin> anything else to discuss? 03:45:31 <hongbin> sounds no more 03:45:40 <hongbin> all, thanks for joining the meeting 03:45:44 <hongbin> #endmeeting