03:00:03 <hongbin> #startmeeting zun
03:00:05 <openstack> Meeting started Tue Mar 14 03:00:03 2017 UTC and is due to finish in 60 minutes.  The chair is hongbin. Information about MeetBot at http://wiki.debian.org/MeetBot.
03:00:06 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
03:00:08 <openstack> The meeting name has been set to 'zun'
03:00:09 <hongbin> #link https://wiki.openstack.org/wiki/Zun#Agenda_for_2017-03-14_0300_UTC Today's agenda
03:00:13 <hongbin> #topic Roll Call
03:00:22 <mkrai_> Madhuri Kumari
03:00:24 <shubhams> shubham
03:00:25 <pksingh> pradeep
03:00:31 <kevinz> kevinz
03:01:08 <hongbin> thanks for joining the meeting mkrai_ shubhams pksingh kevinz
03:01:25 <hongbin> let's get started
03:01:29 <FengShengqin> hi
03:01:40 <hongbin> oh, hey FengShengqin
03:01:45 <hongbin> FengShengqin: thanks for joining
03:01:56 <hongbin> #topic Announcements
03:02:03 <hongbin> 1. Zun will have two presentations in Boston Summit
03:02:09 <hongbin> #link https://www.openstack.org/summit/boston-2017/summit-schedule/events/18591/project-update-zun Project Update - Zun
03:02:14 <hongbin> #link https://www.openstack.org/summit/boston-2017/summit-schedule/events/17988/future-of-containers-in-openstack Future of Containers in OpenStack
03:02:21 <mkrai_> Yayy !!
03:02:22 <pksingh> FengShengqin: You had some problem with IRC in your organization? Good to see you here :)
03:02:25 <hongbin> There are a few other talks that will related to zun
03:02:35 <pksingh> :(
03:02:48 <pksingh> i can't be part of the talk
03:02:51 <FengShengqin> yes,I'm glad too
03:03:23 <hongbin> #topic Review Action Items
03:03:29 <hongbin> #topic Cinder integration (diga)
03:03:34 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/cinder-zun-integration The BP
03:03:51 <hongbin> it looks diga is not here, perhaps we could skip this topic for now
03:04:13 <hongbin> we could revisit it if diga showed up later in the meeting
03:04:50 <hongbin> fyi, it seems he was working on this patch last week
03:04:51 <hongbin> #link https://review.openstack.org/#/c/429943/
03:04:59 <hongbin> #topic Kuryr integration (hongbin)
03:05:05 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/kuryr-integration The BP
03:05:25 <hongbin> i submitted several patches to kuryr-libnetwork, but still waiting for them to be merged
03:06:17 <hongbin> i wanted to get them all merged first before starting to work on the zun side because the implemenation depends on how the patches will be accepted in kuryr side
03:06:46 <pksingh> hongbin: is there any problem in acceptence of those patches in kuryr?
03:06:53 <hongbin> fyi, here are the patches
03:06:55 <hongbin> #link https://review.openstack.org/#/q/project:openstack/kuryr-libnetwork+and+owner:%22Hongbin+Lu+%253Chongbin.lu%2540huawei.com%253E%22
03:07:18 <pksingh> most of them got +2 :)
03:07:20 <hongbin> pksingh: reviewers might have some advices on revising the patches
03:07:29 <kevinz> cool
03:07:54 <hongbin> pksingh: the design might change to address comments from their reviewers
03:08:10 <hongbin> yes, it looks most of them are close to merge
03:08:15 <pksingh> hongbin: ok
03:08:20 <mkrai_> It seems only few are left to be merged
03:08:24 <hongbin> hopefully, i can get them all merged by this week
03:08:34 <pksingh> great !!
03:08:52 <hongbin> ok, any other comment on this topic?
03:08:58 <diga> o/
03:09:03 <diga> sorry got late
03:09:23 <hongbin> diga: hi diga, thanks for joining
03:09:33 <hongbin> #topic Cinder integration (diga)
03:09:39 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/cinder-zun-integration The BP
03:09:41 <diga> hongbin: yeah
03:09:42 <hongbin> diga: ^^
03:10:05 <diga> hongbin: Most of the work is completed, I am facing some problem with Fuxi,
03:10:23 <diga> hongbin: if you have time tomorrow your day time, can you help on it ?
03:10:38 <hongbin> diga: yes, i will try
03:10:44 <mkrai_> diga: What kind of issue?
03:11:15 <diga> hongbin: Fuxi installation & some integration issues are there
03:11:31 <diga> hongbin: will mail you seperately on this,
03:12:00 <pksingh> Shunli: Welcome :)
03:12:07 <Shunli> :)
03:12:18 <diga> hongbin: mostly on integration level, something is messing up in docker volume call to fuxi
03:12:19 <hongbin> Shunli: hey, thanks for joining  !
03:12:34 <hongbin> diga: i see
03:12:39 <Shunli> my pleasure
03:12:59 <diga> hongbin: I will send you details on this, I am not able to trace the issue at docker level
03:13:18 <hongbin> diga: sure, i will try my best to help
03:13:27 <diga> hongbin: thank you
03:13:37 <hongbin> thanks diga
03:13:46 <diga> hongbin: wc!
03:14:07 <hongbin> #topic Introduce host capabilities and cpusets (sudipto)
03:14:12 <hongbin> #link https://review.openstack.org/#/c/427007/ The spec
03:14:27 <hongbin> it looks sudipto is not here right now
03:14:32 <diga> hongbin: I have to join another office meeting at 9, need to leave now
03:14:53 <hongbin> diga: ok, ttyl, thanks again for joining the meeting
03:15:05 <diga> hongbin: np
03:15:37 <hongbin> ok, let's table this topic to next week
03:15:57 <hongbin> next one
03:15:59 <hongbin> #topic Discussion of the image API
03:16:06 <hongbin> #link https://etherpad.openstack.org/p/zun-image-api the etherpad
03:16:19 <hongbin> mkrai_: shubhams : want to drive this topic?
03:16:24 <mkrai_> Ok so I will try to give an update on this
03:16:29 <mkrai_> hongbin: Yes
03:16:49 <mkrai_> We (shubham, pksingh and I) had meeting on this and decided on few things
03:16:57 <mkrai_> pksingh: Please feel free to chime in
03:17:08 <pksingh> mkrai_: go ahead :)
03:17:25 <mkrai_> We decided to have some of the image APIs(POST, GET, DELETE) and all restricted to admin
03:17:48 <mkrai_> For multiple hosts, pulling images should be possible for single host, all hosts or list of hosts separated by comma. This can be done in parallel if optimization is needed
03:18:21 <mkrai_> And we will have the 'host' field in image DB to store the hosts information
03:18:31 <FengShengqin> can anyone help how to use mulit zun-computes?
03:18:59 <mkrai_> And some of the APIs like search can be removed
03:19:28 <mkrai_> That's is all from our discussion
03:19:35 <hongbin> FengShengqin: your questions is about the image api?
03:19:41 <mkrai_> What are teams' opinion on it?
03:19:42 <FengShengqin> no
03:20:04 <hongbin> FengShengqin: we can revisit your question later in the meeting agenda
03:20:15 <FengShengqin> thanks
03:20:48 <hongbin> mkrai_: if the search api is removed, users won't be able to search in dockerhub/glance?
03:21:01 <mkrai_> hongbin: Yes
03:21:23 <hongbin> mkrai_: then they are expacted to use docker/glance cli to do that
03:21:46 <hongbin> mkrai_: which seems to be fine
03:21:51 <mkrai_> hongbin: Yes
03:22:37 <hongbin> mkrai_: the only concern i have is scalability of storing all images X hosts db entries
03:22:59 <hongbin> mkrai_: however, i think it would be fine to get started in this way as proposed
03:23:16 <hongbin> mkrai_: we can revisit it if the scalability is proven to be a problem later
03:23:26 <mkrai_> I think having a single entry with 'host' field a list to store the images should do
03:23:50 <mkrai_> sorry store the host info
03:24:31 <hongbin> perhaps it is better to start with a simpler implementation (host per image)
03:24:55 <hongbin> the optimization like storing list of hosts could be done at the second iteration
03:24:56 <mkrai_> Ok sounds good
03:25:03 <mkrai_> yes
03:25:11 <pksingh> hongbin: +1, it sounds good as of now
03:25:38 <mkrai_> Cool. Thanks pksingh and shubhams for the same :)
03:25:43 <hongbin> mkrai_: pksingh : i have no problem of this approach, it sounds good to me
03:25:54 <hongbin> others, any comment?
03:25:55 <mkrai_> Thanks hongbin
03:26:05 <pksingh> :)
03:26:11 <mkrai_> I will go ahead with the implementation
03:26:47 <hongbin> mkrai_: thanks
03:27:14 <hongbin> mkrai_: shubhams pksingh thanks for all of you to work out the proposal
03:27:48 <hongbin> any other comment before advancing topic?
03:28:12 <hongbin> ok, move on
03:28:14 <hongbin> #topic Introduce container composition
03:28:23 <hongbin> kevinz: want to drive this topic?
03:28:35 <hongbin> #link https://review.openstack.org/#/c/437759/
03:28:39 <kevinz> Yeah ~ Wenzhi and me will work on this topic
03:28:57 <kevinz> And I've uploaded new version of the spec now
03:30:29 <hongbin> the spec looks quite close to remove the wip
03:30:41 <kevinz> In the new spec, we plan to implement capsule as K8S pod and incompatible with docker compose yml file(by setting the kind filed)
03:30:45 <kevinz> field
03:31:16 <hongbin> kind field?
03:31:36 <kevinz> Yeah "kind" field in yaml file.
03:31:53 <hongbin> what this field is going to represent?
03:32:01 <hongbin> docker or pod?
03:32:36 <kevinz> The "kind" can be set to "compose", if we want to praser the docker-compose file
03:32:56 <hongbin> i see
03:33:43 <hongbin> not sure how much this field would help
03:34:10 <hongbin> but that is just my feeling
03:34:11 <mkrai_> Me too. What are the other valid option for field?
03:34:55 <pksingh> may be having different APIs would help more, but not sure
03:35:33 <kevinz> Now just "pod" and "compose", may be add more in the future
03:36:10 <kevinz> Differet "kind" field will call different praser for the yaml
03:36:37 <mkrai_> Ok. I will try to visit the spec today and get more detail.
03:36:44 <yuanying> Does latest docker swarm support docker-compose?
03:36:58 <kevinz> pksingh: Yeah that is another way to implement
03:37:11 <hongbin> yuanying: hey, good to see you here :)
03:37:30 <yuanying> hongbin: hi, me too
03:37:57 <kevinz> yuanying: Not sure, I will check. So IMO, if docker-compose is not important, we will remove its support ?
03:38:21 <yuanying> I guess, docker-compose and pod are different
03:38:50 <yuanying> docker-compose is just a template, right??
03:39:42 <kevinz> Yeah, just a template for multiple dockers. Don't have much functions
03:39:57 <yuanying> yes, but pod has other meaning
03:40:15 <yuanying> it has tightly related conatiners
03:40:51 <yuanying> So, I think we should avoid to think
03:41:11 <hongbin> kevinz: i think yuanying bring up a great point here
03:41:31 <yuanying> to regard the same between pod and docker-compose
03:41:31 <hongbin> compose and pod are very different in implementation and meaning
03:41:44 <kevinz> Yeah, compatible with docker compose file may introduce more problem.
03:41:45 <hongbin> i would rather to pick one instead of picking both
03:41:45 <yuanying> maybe
03:42:24 <kevinz> Right, so I think pod is good to implement now
03:42:31 <hongbin> ok, wfm
03:42:35 <mkrai_> +1 for pod
03:42:40 <yuanying> +1
03:42:48 <pksingh> +1
03:42:51 <kevinz> May be table docker-compose compatible in the future
03:43:06 <hongbin> great
03:43:12 <kevinz> OK Thanks~ I will update the spec for review
03:43:15 <hongbin> it looks we agreed on a direction
03:43:22 <hongbin> kevinz: thanks
03:43:24 <kevinz> :-)
03:43:45 <kevinz> My pleasure ~ Thanks all for the valuable advice
03:43:57 <hongbin> any other comment on this topic?
03:44:27 <hongbin> #topic Manage resource constraints per sandbox
03:44:38 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/resource-management
03:45:11 <hongbin> i wanted to bring this up to see if everyone agree on this proposal
03:45:24 <hongbin> or feel free to bring up any opposing point of view
03:46:13 <hongbin> the idea is very simple: adding --cgroup-parent option to poin to the sandbox (infra container) when lauching the real container
03:46:46 <hongbin> e.g. docker run ... -cgroup-parent <sandbox> ...
03:46:57 <mkrai_> hongbin: Currently there is one container per sandbox. Right?
03:47:05 <hongbin> mkrai_: yes
03:47:19 <mkrai_> And does this mean we can have multiple container inside one sandbox?
03:47:27 <yuanying> Is this related to Pod implementation?
03:47:28 <hongbin> yes
03:47:56 <mkrai_> so it means we are exposing the sandbox containers as well
03:48:08 <hongbin> yuanying: yes, i guess it will be used by kevinz 's proposal
03:48:21 <yuanying> ok
03:48:45 <mkrai_> hongbin: ^
03:48:47 <hongbin> mkrai_: not really, pod doesn't expose infra container either
03:49:10 <mkrai_> Yes so that is what I was thinking
03:49:22 <hongbin> mkrai_: i mean you know there is a pod, but you don't know the infra container, although you can see the resource constraints of the pods
03:50:28 <mkrai_> I can't realise the design now, are you planning to submit any spec for this feature?
03:50:29 <hongbin> mkrai_: you have any specific concern?
03:51:07 <hongbin> mkrai_: i can if i am the one who will take this bp :)
03:51:31 <mkrai_> That will be great :)
03:51:43 <hongbin> if nobody want to take it, i can be the default, then i can work out a spec, that is fine.
03:52:00 <mkrai_> As per my knowledge it is not possible to add containers to pod after it is launched. Right ?
03:52:32 <hongbin> mkrai_: although i am not sure, i think it can
03:52:50 <hongbin> mkrai_: the cgroup should be able to change at runtime
03:53:39 <mkrai_> hongbin: Ok I need to check in k8s
03:54:15 <hongbin> mkrai_: i don't think k8s allow it, but i mean it is technically possible
03:54:36 <mkrai_> yes
03:54:38 <hongbin> ok, let's move to open discussion
03:54:39 <hongbin> #topic Open Discussion
03:55:00 <hongbin> mkrai_: any other concern you have?
03:55:20 <mkrai_> No but I would prefer to have a spec for this, if everyone is ok
03:55:28 <hongbin> mkrai_: ack
03:55:34 <mkrai_> Thanks hongbin
03:55:40 <hongbin> np
03:56:03 <hongbin> FengShengqin: i think you had a question in before, now it is the right time to ask :)
03:57:28 <hongbin> others, any topic to bring up?
03:58:27 <hongbin> seems no, we can end the meeting a little earlier today
03:58:39 <hongbin> all, thanks for joining the meeting, see you next time
03:58:40 <mkrai_> Thank you everyone
03:58:42 <hongbin> #endmeeting