| *** Wenzhi has joined #openstack-zun | 00:43 | |
| *** yanyanhu has joined #openstack-zun | 01:23 | |
| *** yuanying has joined #openstack-zun | 01:24 | |
| *** yuanying has quit IRC | 01:24 | |
| hongbin_ | yanyanhu: Qiming : ping | 01:25 |
|---|---|---|
| hongbin_ | yanyanhu: Qiming eliqiao Wenzhi : i have a proposal which will have a significant change : https://review.openstack.org/#/c/365754/ | 01:26 |
| yanyanhu | hi, hongbin | 01:26 |
| hongbin_ | need your feedback on it | 01:27 |
| Wenzhi | hi hongbin | 01:27 |
| hongbin_ | hey | 01:27 |
| Wenzhi | sure | 01:27 |
| yanyanhu | hongbin_, sure, will check it and leave comments | 01:27 |
| hongbin_ | yanyanhu: thx | 01:27 |
| yanyanhu | so it's about multi-tenant network support? | 01:27 |
| yanyanhu | hongbin_, my pleasure | 01:28 |
| hongbin_ | yes | 01:28 |
| yanyanhu | I see | 01:28 |
| yanyanhu | will take a look at it | 01:28 |
| hongbin_ | I will brief the idea here | 01:28 |
| hongbin_ | the idea is to use nova-docker to provision a container with all the networking | 01:29 |
| hongbin_ | then have zun launch a container and joining the nova-docker container | 01:29 |
| hongbin_ | that is because nova-docker has everything we want: security group, port binding, etc. | 01:30 |
| yanyanhu | so that nova-docker will be the "switch/router" | 01:30 |
| hongbin_ | sort of | 01:30 |
| yanyanhu | but nova docker will be eof in future? | 01:31 |
| yanyanhu | I guess | 01:31 |
| hongbin_ | there are several choice | 01:31 |
| hongbin_ | either we provide something like nova-docker, or claim support for nova-docker | 01:31 |
| hongbin_ | 1. folk nova-docker | 01:31 |
| hongbin_ | 2. contribute to nova-docker | 01:32 |
| yanyanhu | it's ok for zun to rely on nova-docker if it will keep living I think | 01:33 |
| hongbin_ | yanyanhu: my initial idea is to do the port binding by hand, but when i looked into it closer, it requres to copy thousands lines of code from nova, which is undesirable | 01:33 |
| hongbin_ | yanyanhu: will talk to dims about that | 01:34 |
| yanyanhu | hongbin_, yes, if we want to provide similar support in zun, huge effort(or at least code) is unavoidable | 01:34 |
| yanyanhu | hongbin_, that will be great | 01:34 |
| yanyanhu | since he is the maintainer anyway | 01:34 |
| hongbin_ | yes | 01:35 |
| Wenzhi | agree to rely on nova-docker, we can fork it and we can maintain it in zun code base in future | 01:35 |
| hongbin_ | ok | 01:35 |
| Wenzhi | maybe we have a chance to shift to kuryr in future | 01:36 |
| hongbin_ | frankly, i don't think kuryr is right choice | 01:36 |
| hongbin_ | what it does is to that: | 01:37 |
| hongbin_ | 1. users call docker run ... | 01:37 |
| hongbin_ | 2. docker call libnetwork, which call kuryr | 01:37 |
| hongbin_ | 3. kuryr translate the calls to neutron | 01:38 |
| hongbin_ | basically, it is just an API adaptor, it doesn't solve our problems | 01:38 |
| Wenzhi | I see | 01:38 |
| hongbin_ | yanyanhu: btw, the proposal has a side effect that zun will reuse the nova scheduler, resource management, cells, etc. | 01:54 |
| yanyanhu | hongbin_,I thought only the "network bridge" container will be managed by nova? | 01:56 |
| hongbin_ | yanyanhu: it is more than a network bridge | 01:57 |
| yanyanhu | other containers are still managed by Zun | 01:57 |
| hongbin_ | yanyanhu: in container, there is a concept of namespace | 01:57 |
| yanyanhu | hongbin_, yes | 01:57 |
| hongbin_ | yanyanhu: so there is an empty container provisioned by nova | 01:58 |
| yanyanhu | you mean the namespace in k8s? | 01:58 |
| hongbin_ | yanyanhu: then the zun container join the namespaces of the sandbox container | 01:58 |
| yanyanhu | not the linux namespace | 01:58 |
| hongbin_ | yanyanhu: no, linux namespace, i.e. network namespace, IPC namespace | 01:58 |
| yanyanhu | hongbin_, ok, I see | 01:59 |
| hongbin_ | yanyanhu: so the container provisioned by nova already pick up a host | 01:59 |
| yanyanhu | so you mean all other zun containers will run beside this 'empty' container | 02:00 |
| yanyanhu | in the same host | 02:00 |
| hongbin_ | yanyanhu: think of a sandbox like a pod | 02:00 |
| yanyanhu | because they need to be in the same namespace | 02:00 |
| yanyanhu | yes, that will be a pod | 02:00 |
| hongbin_ | yanyanhu: you can have one-to-one mapping between sandbox and zun container | 02:01 |
| hongbin_ | yanyanhu: or one-to-many mapping | 02:01 |
| yanyanhu | ok, sounds interesting | 02:01 |
| hongbin_ | yanyanhu: however, the empty container is scheduled by nova, cpu/memory is specified in nova | 02:01 |
| hongbin_ | yanyanhu: what zun needs to do is just run a container in a sandbox | 02:02 |
| hongbin_ | yanyanhu: nova does everything for the sandbox (networking, resources management, storage, etc.) | 02:02 |
| yanyanhu | then the 'resource requirement' for that 'special' container actually depends on the pod resource demand | 02:03 |
| hongbin_ | yes | 02:03 |
| hongbin_ | so, pod has resource constraint, zun container don't have | 02:03 |
| hongbin_ | or i should say zun container depends on its pod | 02:04 |
| yanyanhu | from service deployment perspective, it makes sense to let pod become the primitive | 02:04 |
| hongbin_ | yes | 02:05 |
| yanyanhu | but there could be case user wants to apply for a single container | 02:05 |
| hongbin_ | then create a pod with one container | 02:05 |
| yanyanhu | then the map will be one-one? | 02:05 |
| hongbin_ | yes | 02:05 |
| hongbin_ | actually, we can do that: zun create <container> | 02:06 |
| yanyanhu | but we still need to expose interface for "container" management, right | 02:06 |
| yanyanhu | not just pod management | 02:06 |
| hongbin_ | it creates a sandbox and a container | 02:06 |
| hongbin_ | could be. the api could be disucssed further | 02:06 |
| hongbin_ | my initial thinking is | 02:06 |
| hongbin_ | 1. zun create ... : create a sandbox and a container | 02:07 |
| hongbin_ | 2. zun create --sandbox ... : create a container and join a existing sandbox | 02:07 |
| hongbin_ | however, pod management is proxyed to nova api calls | 02:08 |
| yanyanhu | hmm, maybe we should hide sandbox container from enduser? | 02:08 |
| * hongbin_ has a feeling that the scope of the proposal is more than networking | 02:09 | |
| yanyanhu | if user want to create a new container which will join an existing network/storage, we handle the sandbox finding/joining at background | 02:09 |
| yanyanhu | and make this progress transparent to user | 02:09 |
| hongbin_ | yes, that can be done | 02:09 |
| yanyanhu | since sandbox is not a conception user should care about :) | 02:10 |
| *** itzdilip has joined #openstack-zun | 02:10 | |
| hongbin_ | yes it might introduce extra complexity | 02:10 |
| yanyanhu | hongbin_, that's true | 02:11 |
| hongbin_ | yanyanhu: for the nova-docker, i might not care. if it is not nova-docker, it needs to be a nova driver | 02:12 |
| hongbin_ | yanyanhu: since we want to reuse everything in nova | 02:12 |
| yanyanhu | understand | 02:12 |
| hongbin_ | ok | 02:12 |
| hongbin_ | will bring this proposal up at the team meeting | 02:13 |
| hongbin_ | to get more feedback | 02:13 |
| yanyanhu | ok | 02:13 |
| hongbin_ | will leave for a while, then come back to run the meeting | 02:14 |
| yanyanhu | ok, ttyl | 02:14 |
| *** hongbin_ has quit IRC | 02:49 | |
| *** hongbin_ has joined #openstack-zun | 02:50 | |
| hongbin_ | all, team meeting will start in about 9 minutes at #openstack-meeting channel | 02:51 |
| *** yuanying has joined #openstack-zun | 02:52 | |
| *** shu-mutou-AFK is now known as shu-mutou | 02:55 | |
| *** yuanying has quit IRC | 02:55 | |
| *** vikasc has quit IRC | 02:56 | |
| *** yuanying has joined #openstack-zun | 02:57 | |
| *** vikasc has joined #openstack-zun | 02:57 | |
| *** yuanying has quit IRC | 03:00 | |
| *** sudipto has joined #openstack-zun | 03:00 | |
| *** sudipto_ has joined #openstack-zun | 03:00 | |
| *** yuanying has joined #openstack-zun | 03:00 | |
| *** shubhams has joined #openstack-zun | 03:01 | |
| *** janki has joined #openstack-zun | 03:09 | |
| *** yuanying has quit IRC | 03:18 | |
| *** yuanying has joined #openstack-zun | 03:20 | |
| *** adisky has joined #openstack-zun | 03:47 | |
| *** janki has quit IRC | 03:51 | |
| *** yuanying has quit IRC | 04:00 | |
| hongbin_ | need to leave now. see you folks | 04:00 |
| *** hongbin_ has quit IRC | 04:01 | |
| *** yuanying has joined #openstack-zun | 04:02 | |
| *** sudipto_ has quit IRC | 04:06 | |
| *** sudipto has quit IRC | 04:06 | |
| *** shu-mutou is now known as shu-mutou-AFK | 04:12 | |
| *** sudipto has joined #openstack-zun | 04:53 | |
| *** sudipto_ has joined #openstack-zun | 04:53 | |
| *** yasemin has joined #openstack-zun | 05:04 | |
| *** chandankumar has joined #openstack-zun | 05:34 | |
| *** janki has joined #openstack-zun | 05:45 | |
| *** sudipto has quit IRC | 05:47 | |
| *** sudipto_ has quit IRC | 05:47 | |
| *** sudipto has joined #openstack-zun | 05:53 | |
| *** sudipto has quit IRC | 07:05 | |
| *** mikelk has joined #openstack-zun | 07:59 | |
| *** Wenzhi has quit IRC | 08:04 | |
| *** sudipto has joined #openstack-zun | 08:34 | |
| *** shubhams has quit IRC | 08:53 | |
| *** chandankumar has quit IRC | 09:19 | |
| *** chandankumar has joined #openstack-zun | 09:25 | |
| *** mfedosin has joined #openstack-zun | 09:25 | |
| *** yanyanhu has quit IRC | 10:01 | |
| *** chandankumar1 has joined #openstack-zun | 10:35 | |
| *** chandankumar has quit IRC | 10:39 | |
| *** yasemin has quit IRC | 10:47 | |
| *** chandankumar1 is now known as chandankumar | 10:54 | |
| *** yasemin has joined #openstack-zun | 11:00 | |
| *** mfedosin has quit IRC | 11:37 | |
| *** mfedosin has joined #openstack-zun | 11:51 | |
| *** sudipto has quit IRC | 11:55 | |
| *** vikasc has quit IRC | 12:47 | |
| *** sudipto has joined #openstack-zun | 12:51 | |
| *** vikasc has joined #openstack-zun | 13:00 | |
| *** vikasc has quit IRC | 13:43 | |
| *** vikasc has joined #openstack-zun | 13:55 | |
| *** chandankumar has quit IRC | 14:00 | |
| *** chandankumar has joined #openstack-zun | 14:01 | |
| *** janki has quit IRC | 14:26 | |
| *** mikelk has quit IRC | 14:49 | |
| *** janki has joined #openstack-zun | 15:13 | |
| *** sudipto has quit IRC | 15:31 | |
| *** tbh has joined #openstack-zun | 15:52 | |
| *** manikanta_tadi has joined #openstack-zun | 15:59 | |
| *** chandankumar has quit IRC | 16:10 | |
| *** manikanta_tadi has quit IRC | 16:34 | |
| *** tbh has quit IRC | 16:59 | |
| *** janki has quit IRC | 17:08 | |
| *** harlowja has joined #openstack-zun | 17:38 | |
| *** mfedosin has quit IRC | 18:40 | |
| *** vikasc has quit IRC | 19:42 | |
| *** vikasc has joined #openstack-zun | 19:44 | |
| *** flwang1 has quit IRC | 20:21 | |
| openstackgerrit | Hongbin Lu proposed openstack/zun: Fix cannot delete container in Error status https://review.openstack.org/365290 | 23:07 |
| openstackgerrit | Hongbin Lu proposed openstack/zun: Add more parameters for container create https://review.openstack.org/358378 | 23:14 |
| *** hongbin has quit IRC | 23:17 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!