03:00:01 <hongbin> #startmeeting zun
03:00:02 <openstack> Meeting started Tue Aug 22 03:00:01 2017 UTC and is due to finish in 60 minutes.  The chair is hongbin. Information about MeetBot at http://wiki.debian.org/MeetBot.
03:00:02 <hongbin> #link https://wiki.openstack.org/wiki/Zun#Agenda_for_2017-08-22_0300_UTC Today's agenda
03:00:03 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
03:00:05 <openstack> The meeting name has been set to 'zun'
03:00:07 <hongbin> #topic Roll Call
03:00:11 <spn> o/
03:00:11 <Namrata> Namrata
03:00:14 <mkrai> Madhuri
03:00:17 <pksingh> o/
03:00:34 <kevinz> o/
03:00:47 <Shunli> o/
03:00:54 <hongbin> thanks for joining the meeting spn Namrata mkrai pksingh kevinz Shunli
03:01:00 <hongbin> ok, let's get started
03:01:05 <hongbin> #topic Announcements
03:01:10 <hongbin> 1. Zun 0.2.0 release
03:01:15 <hongbin> #link https://review.openstack.org/#/c/494541/
03:01:26 <hongbin> the stable/pike branch is cute based on this release
03:01:40 <hongbin> so this is the official pike release
03:01:58 <hongbin> thanks all for working hard in this cycle to deliver it
03:02:28 <hongbin> from now no, bug fixes could be back port to stable/pike branch
03:02:41 <hongbin> however, it needs to be fixed in master branch first
03:02:55 <hongbin> any comment on this ?
03:03:10 <mkrai> Thank you for your efforts
03:03:29 <hongbin> mkrai: np, thanks all to deliver this release :)
03:04:06 <hongbin> seems no more comment, continue
03:04:07 <pksingh> i could not contribute much in this release, hope to contribute much in next one :)
03:04:28 <hongbin> pksingh: ack, thanks for your contribution :)
03:04:34 <hongbin> #topic Review Action Items
03:04:41 <hongbin> 1. hongin create a bp for api version negoiation (DONE)
03:04:46 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/api-version-negotiation
03:05:04 <hongbin> i tried to outline the requirements in the bp
03:05:29 <hongbin> please feel free to review this bp and take it if you interest
03:05:50 <hongbin> note: you can ask a core to assign a bp to you if you couldn't assign it yourself
03:06:09 <hongbin> ok, move on
03:06:20 <hongbin> #topic Cinder integration
03:06:39 <hongbin> for this one, i got a WIP patch up for review
03:06:52 <hongbin> #link https://review.openstack.org/#/c/473115/
03:07:11 <hongbin> it is a huge patch, i am going to break it down into several smaller patches
03:07:14 <pksingh> huge patch :)
03:07:20 <hongbin> :)
03:07:57 <hongbin> the idea is to call cinder API to connect to the volume, them bindmount it to the container
03:08:31 <hongbin> the cinder attach workflow is a bit complicated, and needs to handle a lot of exceptions
03:09:07 <hongbin> i tried to break everything into module, hope it is more clear
03:09:15 <mkrai> is the patch functional?
03:09:31 <hongbin> mkrai: not yet, just for a preview
03:09:46 <mkrai> Ok wanted to try out before reviewing
03:09:49 <hongbin> mkrai: if i removed the WIP, then the patch is funtional
03:10:04 <mkrai> hongbin: Ok
03:10:21 <hongbin> mkrai: will let you know once it is ready to try
03:10:31 <mkrai> hongbin: Thank you
03:10:38 <hongbin> mkrai: np
03:10:54 <hongbin> i will continue to work on the patch this week
03:10:59 <hongbin> that is all from me
03:11:08 <hongbin> any other question/comment?
03:12:04 <hongbin> seems no, move on to the next item in agenda
03:12:08 <hongbin> #topic Introduce container composition (kevinz)
03:12:15 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/introduce-compose
03:12:18 <hongbin> kevinz: ^^
03:12:54 <kevinz> This week I'm struggling with capsule test case in API side.
03:13:58 <kevinz> Because it introduce the experimental endpoint, so thing not very easy to write test case. However, I will continue to finish this today or tomorrow
03:14:43 <hongbin> kevinz: unit tests ? or tempest tests?
03:14:55 <kevinz> Unit test
03:15:00 <hongbin> i see
03:15:48 <kevinz> I will ping you for some help offline if possible :-)
03:16:00 <hongbin> kevinz: sure
03:16:26 <hongbin> kevinz: in addition, i wanted to mention that there are progress for this bp last week
03:16:33 <pksingh> kevinz feel free to ping me too, if i can help you somehow
03:16:51 <hongbin> kevinz: because we just merged a (huge) patch for capsule-create
03:16:57 <kevinz> hongbin: pksingh: Thanks a lot
03:17:15 <kevinz> hongbin: Yes, a good progress
03:17:30 <hongbin> #link https://review.openstack.org/484602
03:17:47 <hongbin> there are several other patches are merged as well
03:18:13 <hongbin> there is one to configure the devstack script to add the experimental endpoint
03:18:25 <hongbin> there is a client patch that add support for the capsule api
03:18:32 <kevinz> https://review.openstack.org/#/c/480455/
03:18:36 <hongbin> those are great patches :)
03:18:59 <hongbin> kevinz: thanks for your hard work on this feature :)
03:19:05 <mkrai> kevinz: Thank you :)
03:19:14 <kevinz> hongbin: mkrai: My pleasure
03:19:35 <kevinz> BTW, this patch need another review https://review.openstack.org/#/c/480455/
03:19:50 <kevinz> So that it can be merge in client side
03:20:23 <hongbin> ack, thanks kevinz
03:20:29 <Namrata> I will  go through it and review
03:20:47 <kevinz> Namrata: Thx in advance:-)
03:21:15 <hongbin> ok, move on
03:21:18 <hongbin> #topic NFV use cases (lakerzhou)
03:21:38 <hongbin> lakerzhou seems not here today
03:21:57 <hongbin> for this bp, shunli has several pci patches that are merged last week
03:22:29 <Shunli> will go on working on the database and objects
03:22:51 <hongbin> Shunli: ack
03:23:38 <hongbin> Shunli: after the databae and objects is done, what will be the next step, the scheduler?
03:23:49 <hongbin> or the compute?
03:24:08 <Shunli> scheduler
03:24:15 <hongbin> i see
03:24:28 <Shunli> oh, sorry.maybe the compute first
03:24:43 <Shunli> first need the pci tracker to collect the pci info
03:25:07 <hongbin> second, report the pci info to the scheduler (i guess)
03:25:23 <Shunli> yes.
03:25:28 <hongbin> get it
03:25:40 <hongbin> sounds like a plan
03:25:54 <hongbin> Shunli: thanks shunli for working on this
03:26:03 <Shunli> np. my pleasure
03:26:09 <hongbin> :)
03:26:20 <hongbin> all, any comment on this topic?
03:26:21 <Shunli> :)
03:27:17 <hongbin> ok, next topic
03:27:23 <hongbin> #topic Zun connector for k8s
03:27:29 <hongbin> #link https://blueprints.launchpad.net/zun/+spec/zun-connector-for-k8s
03:27:42 <hongbin> we have an etherpad up for this one
03:27:46 <hongbin> #link https://etherpad.openstack.org/p/zun-connector-k8s
03:28:04 * hongbin is reviewing the etherpad
03:29:07 <Shunli> hongbin: I refered last team meeting is about https://kubernetes.io/docs/concepts/api-extension/apiserver-aggregation/
03:29:54 <hongbin> Shunli: go ahead
03:29:57 <Shunli> is it possible to integrate zun with k8s like integrate zun with heat, define some customed resource type in k8s for zun.
03:30:50 <Shunli> provide zun api as k8s's third party api using the apiserver-aggregation.
03:31:06 <hongbin> i think this is possible technically
03:31:38 <Shunli> I just think it's a easy and feasible way to integrate zun with k8s
03:32:23 <mkrai> I think the intention is not to implement anything in k8s
03:32:51 <mkrai> but provide a similar implementation in zun so that it can be consumed by k8s
03:33:14 <mkrai> Shunli: hongbin WDYT?
03:33:21 <pksingh> and i think we discussed that idea in past too
03:33:29 <kevinz> As I see, modify code in K8S is very hard and out of control
03:33:43 <hongbin> mkrai: i agree with you comment in general
03:33:46 <mkrai> kevinz: Right, that's my point as well :)
03:34:38 <Shunli> yes, modify code in k8s is some kinds of out of control.
03:35:38 <hongbin> in addition, i see it this way:
03:35:56 <hongbin> Shunli's idea is about extend k8s API to implement a Zun API
03:36:27 <hongbin> The ACI connector is about bridge k8s to Zun (with k8s API)
03:36:40 <hongbin> the intention are very different
03:37:38 <kevinz> I will check it. That's a new idea
03:38:42 <mkrai> Sorry I didn't get the first point
03:38:54 <mkrai> Does it mean we extend the k8s API in Zun?
03:39:11 <mkrai> user -> zun -> k8s ?
03:39:23 <hongbin> no, shunli proposal is a k8s api extension for zun
03:40:00 <hongbin> so in k8s api, there will be pod, service , ... , then there is an extension that provide "container", which is a zun container
03:40:12 <mkrai> user->k8s->zun. Right?
03:40:19 <hongbin> yes
03:40:34 <mkrai> hongbin: That seems very difficult to get through k8s community :)
03:40:49 <hongbin> mkrai: yes, it is
03:41:01 <hongbin> possibly, it needs to be out-of-tree
03:41:45 <hongbin> frankly, i am not sure if k8s users interest in a "container" api if there is already a "pod" that they are familiar with
03:41:58 <mkrai> hongbin: I agree completely
03:42:16 <mkrai> container endpoint in k8s wouldn't make sense
03:42:17 <hongbin> Shunli: i see the idea, but not sure if it will work , that is my comment :)
03:42:32 <Shunli> hongbin:
03:43:00 <Shunli> i'm creating the paste, maybe we can discuss if it possible
03:43:15 <hongbin> Shunli: sure
03:43:46 <hongbin> Shunli: in addition, feel free to list your idea in the etherpad (as option #3 in design): https://etherpad.openstack.org/p/zun-connector-k8s
03:44:16 <Shunli> ok, thx.
03:44:22 <hongbin> #topic Open Discussion
03:44:50 <hongbin> any other topic to bring up ?
03:44:51 <mkrai> hongbin: Can I go ahead with the clear container patches?
03:45:00 <hongbin> mkrai: feel free
03:45:08 <mkrai> hongbin: Thank you
03:45:16 <kevinz> mkrai: I have a question about clear container
03:45:23 <mkrai> kevinz: sure
03:45:48 <hongbin> mkrai: if you feel like, i can allocate a topic for clear container in the meeting
03:45:50 <kevinz> mkrai: does it mature to commercial use case?
03:46:00 <mkrai> hongbin: Sure
03:46:05 <hongbin> mkrai: ack
03:46:22 <mkrai> kevinz: I guess yes :)
03:46:32 <mkrai> It is integrated with k8s, swarm as well
03:46:40 <kevinz> mkrai: ACK
03:46:47 <spn> In fact it makes more sense using CC with all its security around container
03:46:59 <hongbin> spn: ++
03:47:29 <mkrai> spn: yes :)
03:47:32 <hongbin> for example, in public cloud, strong-isolated container like CC is a must
03:48:04 <hongbin> so clear container integration is very import for zun
03:48:20 <spn> but definately we should have an option of running regular containers and clear containers on same computes
03:48:46 <hongbin> i see
03:49:10 <mkrai> spn: We will have that support
03:49:40 <spn> mkrai: great :)
03:50:01 <hongbin> have both in the same compute node is a bit challenging, but i believe it can be done
03:50:15 <mkrai> Bdw I and spn will have a session at sydney on clear container, if interested please join :)
03:51:03 <mkrai> hongbin: can you tell the issues if possible?
03:51:33 <hongbin> mkrai: i don't know what will be the issues, the challenging is like to ask nova to support two hypervisors in the same node :)
03:51:57 <spn> probably run multiple docker daemons on different port numbers one with coe runtime
03:51:58 <hongbin> perhaps , it will be easy, i am just not sure
03:52:44 <mkrai> hongbin: ack. I will note this point and see if there's any issue
03:53:13 <hongbin> i see
03:53:31 <mkrai> spn: I am not sure of that now. But seems that can be an option.
03:53:57 <spn> mkrai: yup!
03:54:08 <mkrai> hongbin: spn Thanks
03:54:56 <hongbin> all, any other topic or question?
03:55:16 <mkrai> None from me
03:55:54 <hongbin> all, thanks for joining the meeting, see you next time :)
03:55:58 <hongbin> #endmeeting