*** openstack has joined #openstack-zun | 01:24 | |
*** shu-mutou has joined #openstack-zun | 01:33 | |
*** yanyanhu has joined #openstack-zun | 01:50 | |
*** sheel has joined #openstack-zun | 02:48 | |
*** yuanying has quit IRC | 02:50 | |
*** klkumar has joined #openstack-zun | 02:58 | |
*** yuanying has joined #openstack-zun | 03:47 | |
*** sudipto has joined #openstack-zun | 04:42 | |
*** flwang1 has quit IRC | 05:00 | |
*** irenab has joined #openstack-zun | 05:07 | |
*** chandankumar has joined #openstack-zun | 05:49 | |
*** sudipto has quit IRC | 06:54 | |
*** flwang1 has joined #openstack-zun | 07:26 | |
*** klkumar has quit IRC | 08:21 | |
*** shu-mutou has quit IRC | 08:25 | |
openstackgerrit | Merged openstack/higgins: Fix error in quickstart https://review.openstack.org/334847 | 08:27 |
---|---|---|
*** klkumar has joined #openstack-zun | 08:35 | |
*** flwang1 has quit IRC | 08:39 | |
openstackgerrit | Merged openstack/higgins: Add Higgins Service Controller https://review.openstack.org/325611 | 08:51 |
*** flwang1 has joined #openstack-zun | 09:26 | |
*** sudipto has joined #openstack-zun | 09:27 | |
*** yanyanhu has quit IRC | 10:04 | |
*** mkrai_ has joined #openstack-zun | 10:20 | |
*** sudipto has quit IRC | 10:36 | |
*** sudipto has joined #openstack-zun | 10:53 | |
*** mkrai_ has quit IRC | 11:07 | |
*** sudipto has quit IRC | 11:14 | |
*** mkrai has joined #openstack-zun | 11:28 | |
*** irenab has quit IRC | 11:29 | |
*** irenab has joined #openstack-zun | 11:40 | |
*** chandankumar has quit IRC | 11:50 | |
*** chandankumar has joined #openstack-zun | 11:51 | |
*** klkumar has quit IRC | 12:13 | |
*** mkrai has quit IRC | 12:13 | |
*** chandankumar has quit IRC | 12:45 | |
*** sudipto has joined #openstack-zun | 12:54 | |
*** hongbin has joined #openstack-zun | 13:34 | |
hongbin | sudipto: there? | 13:35 |
sudipto | hongbin, hello | 13:35 |
hongbin | hey | 13:35 |
hongbin | A good time to discuss the API design ? | 13:35 |
sudipto | Can you give me like 15 mins? Working on a set of patches in nova (deadline for feature freeze is in another 2 days)... else if you are free later in your day, i am going to be around. | 13:36 |
hongbin | sure. Just ping me after | 13:36 |
sudipto | hongbin, done! | 13:37 |
hongbin | done? 15 mins.... | 13:37 |
sudipto | hongbin, sure if you prefer that or if you are free sometime in another couple of hours, that works too. | 13:40 |
hongbin | I can start now | 13:41 |
sudipto | hongbin, ok sure. | 13:41 |
hongbin | sudipto: I sent an email yesterday to summarize the design | 13:41 |
hongbin | That is my understanding of the agreement. Hope I didn't misunderstand something | 13:42 |
hongbin | We can start the discussion based on that | 13:42 |
sudipto | hongbin, yeah - i went through it - i think that's a good summary. I wanted to reply saying - we may want to start with the reference architecture to begin with - to find that minimal overlap of APIs you talked about between COEs? | 13:42 |
hongbin | ok | 13:43 |
sudipto | do you agree? | 13:43 |
hongbin | I think it is good to start with basic | 13:43 |
*** Qiming has quit IRC | 13:44 | |
hongbin | sure with reference architecture? | 13:44 |
hongbin | sure | 13:44 |
*** Qiming has joined #openstack-zun | 13:44 | |
hongbin | s/sure/start/ | 13:44 |
hongbin | what do you think about the reference architecture? | 13:44 |
sudipto | yeah i think we will otherwise struggle to map what we want to overlap between let's say K8s and swarm | 13:46 |
*** klkumar has joined #openstack-zun | 13:46 | |
hongbin | sudipto: Yes. I wonder if you have any idea about the reference architucture | 13:47 |
hongbin | sudipto: what the architecture should look like? | 13:48 |
sudipto | hongbin, yeah - been thinking about it. I guess it's important to define what we want...and from your mail - i have the understanding that we want something like kubernetes. right? | 13:48 |
hongbin | sudipto: Sort of. We want a wrapper of kubernetes | 13:49 |
sudipto | hongbin, as in Kubernetes is wrapper is the integration to the COE part...however - then we also want to call docker APIs as a part of ZUN (and i thought we want to approach this like kubernetes) | 13:50 |
hongbin | Oh, the reference implementation | 13:50 |
hongbin | That is different | 13:50 |
hongbin | https://etherpad.openstack.org/p/zun-containers-service-api | 13:51 |
hongbin | sudipto: I think the reference implementation should not look like kubernetes, otherwise, it will have a lot of work :) | 13:51 |
sudipto | hongbin, heh, got it now. I thought we wanted to approach it that way. | 13:51 |
hongbin | sudipto: maybe just a container abstraction is enough | 13:51 |
hongbin | ok | 13:52 |
sudipto | hongbin, so let's start with an etherpad for the API design? | 13:52 |
hongbin | this one? https://etherpad.openstack.org/p/zun-containers-service-api | 13:53 |
sudipto | This seems almost done, from the point of view of doing a docker native API support, no? | 13:53 |
sudipto | or do you want to run through this once? | 13:54 |
*** chandankumar has joined #openstack-zun | 13:54 | |
hongbin | No. I mean the API desing is there | 13:54 |
hongbin | sudipto: OK. Maybe we need to discuss the API things | 13:55 |
sudipto | I think we should go with docker registry APIs as opposed to glance for images to begin with - what do you say? | 13:55 |
sudipto | since glance support may take sometime to get built. | 13:56 |
hongbin | Glance already support docker | 13:56 |
sudipto | hongbin, yeah it does but it does not support anything beyond storing the docker gz file. | 13:56 |
hongbin | (flwang has pointed that out in the meeting | 13:56 |
hongbin | sudipto: Yes, it doesn't support layer of docker iamges | 13:57 |
sudipto | I thought we discussed with flwang and he agreed that the support is only that we store the gz files. That's what i see in the glance documentation and of whatever i have used. | 13:57 |
hongbin | Yes, I guess gz files is enough for us to start? | 13:58 |
sudipto | which means, that we anyway ask the user to use something like "docker pull" and then "docker save" | 13:58 |
hongbin | I see | 13:58 |
hongbin | If we use glance, the workflow is something like this: | 13:58 |
hongbin | docker pull -> docker export -> glance create -> container create | 13:59 |
hongbin | docker pull -> docker export produce a tar file | 13:59 |
hongbin | glance image-create upload the tar file teh glance | 14:00 |
sudipto | https://wiki.openstack.org/wiki/Docker --> Using Nova-Docker | 14:00 |
hongbin | zun container-create --image <GLANCE_IMAGE> | 14:00 |
sudipto | this documents the glance support today | 14:00 |
sudipto | yeah that could also be done for now. | 14:00 |
hongbin | OK | 14:01 |
hongbin | That is the short term solution | 14:01 |
sudipto | however, glance doesn't support a "docker commit" | 14:01 |
sudipto | which means that the user again goes back to the pull/save workflow in order to use the image. | 14:01 |
hongbin | sudipto: I see | 14:02 |
hongbin | THen, how about create a command "zun container-commit" | 14:02 |
hongbin | That upload commit the image and upload it to glance | 14:02 |
sudipto | that sounds interesting :) | 14:03 |
sudipto | so basically this command under the hood would do a glance registry upload - then pull back the image down to glance? | 14:03 |
hongbin | yes | 14:03 |
hongbin | From end-users point of view, it should be similiar to docker commit | 14:04 |
sudipto | however, this image is just an increment from the last one ... | 14:04 |
hongbin | I see | 14:04 |
sudipto | we can however, do this workflow and have a TODO to glance to do incremental saves - what say? | 14:05 |
hongbin | yes glance needs to support that | 14:06 |
hongbin | If not, we need a work around | 14:06 |
sudipto | Glance or Glare - whichever would bring the image support. Nikhil (Glance PTL) - is interested in this and so is mfedosin (Glare lead)...I will have a talk with them... | 14:06 |
sudipto | flwang, is the best person in any case :) | 14:07 |
hongbin | flwang1: ^^ | 14:07 |
sudipto | ok let's move on then? | 14:08 |
hongbin | k | 14:08 |
hongbin | BTW, I haven't thought about the image support thoughfully. That worthes a spec to talk about it | 14:09 |
sudipto | yeah i was just jotting it down on the etherpad then realize there might be some optimizations done - since we would take a round about route to get to glance... | 14:10 |
hongbin | maybe cache in local, or similiar optimization | 14:11 |
sudipto | yeah | 14:12 |
hongbin | sudipto: For the API design, you said you want one set of APIs instead of two set ? | 14:12 |
sudipto | hongbin, nope. I thought i read it that way. | 14:12 |
sudipto | in your mail | 14:12 |
hongbin | ok | 14:12 |
hongbin | In the email, I explained there is two set of APIs | 14:13 |
hongbin | The first set for COEs, the other set for container runtimes | 14:13 |
sudipto | but i think you were clear about 2 sets of API. My perception was, we are not really supporting the docker native APIs at all, instead supporting a cluster management from OpenStack - and that's what the reference architecture was for the COEs. | 14:13 |
sudipto | yeah - that makes it clear... | 14:13 |
hongbin | ok. Good that this is clear | 14:14 |
hongbin | sudipto: did you see any pitfall in this design? | 14:15 |
* sudipto seeks 10 mins from hongbin to work on a review comment. Sorry. | 14:15 | |
hongbin | np | 14:15 |
*** klkumar has quit IRC | 14:17 | |
*** sheel has quit IRC | 15:15 | |
*** sheel has joined #openstack-zun | 15:49 | |
*** klkumar has joined #openstack-zun | 16:03 | |
*** klkumar has quit IRC | 17:31 | |
*** harlowja has joined #openstack-zun | 17:37 | |
harlowja | hongbin i'm around when u around | 17:37 |
*** sudipto has quit IRC | 17:44 | |
hongbin | harlowja: never mind. I asked the sig-openstack | 17:52 |
harlowja | k | 17:52 |
*** sheel has quit IRC | 18:55 | |
*** chandankumar has quit IRC | 19:38 | |
*** flwang1 has quit IRC | 20:36 | |
*** hongbin has quit IRC | 23:34 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!