03:00:18 #startmeeting zun 03:00:19 Meeting started Tue Jan 24 03:00:18 2017 UTC and is due to finish in 60 minutes. The chair is hongbin. Information about MeetBot at http://wiki.debian.org/MeetBot. 03:00:20 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 03:00:22 The meeting name has been set to 'zun' 03:00:25 #link https://wiki.openstack.org/wiki/Zun#Agenda_for_2017-01-24_0300_UTC Today's agenda 03:00:29 #topic Roll Call 03:00:31 Namrata 03:00:32 shubham 03:00:34 kevinz 03:01:24 thanks for joining the meeting Namrata shubhams kevinz 03:01:43 Lakerzhou 03:01:52 Madhuri Kumari 03:01:54 i knew these weeks are lunar new year, so some team members might not be able to join 03:02:02 o/ 03:02:02 hey lakerzhou mkrai 03:02:06 hi diga 03:02:15 hongbin: Hi 03:02:18 ok, let's start 03:02:18 Hi Jong in 03:02:24 #topic Announcements 03:02:29 i have one 03:02:51 for zun cores, please vote on the proposal: http://lists.openstack.org/pipermail/openstack-dev/2017-January/110724.html 03:02:55 #link http://lists.openstack.org/pipermail/openstack-dev/2017-January/110724.html 03:03:08 that is it from my side 03:03:15 anyone else has any announcement? 03:03:31 #topic Review Action Items 03:03:32 none 03:03:37 #topic Cinder integration (diga) 03:03:41 #link https://blueprints.launchpad.net/zun/+spec/cinder-zun-integration The BP 03:03:46 #link https://review.openstack.org/#/c/417747/ The design spec 03:03:50 diga: ^^ 03:04:01 yes 03:04:19 hongbin: I started the development yesterday 03:04:32 diga: cool 03:04:44 hongbin: By end of the week, will submit the first PS 03:05:12 Hello 03:05:16 diga: awesome, looking forward to your patch 03:05:23 pksingh: hey 03:05:29 hongbin: :) 03:05:53 sorry, i got litle late 03:05:58 one thing i am going to add, the design spec of cinder integration has merged 03:06:07 #link https://review.openstack.org/#/c/417747/ The design spec 03:06:08 hongbin: yes 03:06:21 however, let diga know if there is something that you have a concern 03:06:25 pksingh: np 03:06:58 any other question for diga ? 03:07:26 ok, advance topic 03:07:26 hongbin: No, I think I am good to go now on the submitted spec 03:07:32 thanks diga 03:07:38 #topic Support interactive mode (kevinz) 03:07:41 hongbin: welcome! 03:07:45 #link https://blueprints.launchpad.net/zun/+spec/support-interactive-mode The BP 03:07:50 #link https://review.openstack.org/#/c/396841/ The design spec 03:07:52 kevinz: ^^ 03:08:02 Hi 03:08:42 I have uploaded some wip patches both in server side and client side. 03:09:07 kevinz: you have some links? 03:09:10 https://review.openstack.org/#/c/422525/ https://review.openstack.org/#/c/417681/ 03:09:31 #link https://review.openstack.org/#/c/422525/ 03:09:36 #link https://review.openstack.org/#/c/417681/ 03:11:03 kevinz: perhaps you could briefly talk about what is the current status of this bp and what these patches are about 03:11:15 BTW, both the client and server side need more test cases 03:11:31 will find some time to review these patches today 03:12:14 OK. The client side is working for websocket client, user will do the "attach" command , get a websocket link from API server. 03:13:09 The client side use select poll mechanism and some interpt to deal with user input , output and stderr 03:14:29 kevinz: cool 03:14:36 In server side, there are two API, one is about "attach", mainly to get the Docker daemon version. The other is "resize", to do the resize tty function in the compute node 03:15:06 That's the briefly introduction about these two patch :-) 03:15:28 kevinz: thanks for the introduction 03:15:42 kevinz: is there any road blocker so far for implementing this bp? 03:15:57 kevinz: e.g. the docker-py version, etc. 03:17:16 Hongbin: Just one problem, websocket link need the compute node ip, or hostname 03:17:38 kevinz: use the host_ip in the config file will work? 03:18:33 hongbin: Does API server know all the compute nodes' IP? 03:19:18 kevinz: i am not sure exactly, it seems no 03:19:50 I konw API server can search from DB to get all computes node's hostname, but I'm not sure this "hostname" can work in websocket link, I will check it 03:19:52 kevinz: i think the compute host needs to return the ip address 03:20:36 hongbin: OK, that is a solution, I will add to the attach function. 03:20:42 Thanks 03:20:56 kevinz: alternatively, there is a 'host' field in container, check that as well 03:21:04 kevinz: thanks kevinz 03:21:20 all, any other question for kevinz ? 03:21:24 Yeah ,I will check this. 03:21:57 ok, next topic 03:22:00 #topic How to expose CPU configurations for containers 03:22:30 it looks sudipta is not here, but he discussed with me in before about that 03:22:57 both I and he agree to expose the vcpu as the resource constraint 03:23:06 for simplicity 03:23:43 at the backend, we tranaslate vcpu into docker specific config (e.g. cpu_period, cpu_share, etc.) 03:24:15 seems interesting 03:24:26 later, sudipta will propose a feature to pin to container to specific cpu cores 03:24:46 all, thoughts? 03:25:31 i did not get much time last week, will look more into it 03:25:47 I guess we need a spec for this 03:25:59 mkrai: +1 03:26:04 mkrai: ack 03:26:32 i think we need to create a bp first 03:26:45 #action hongbin create a bp for exposing container cpu resource 03:27:15 ok, it looks that is all the discussion about this topic, anything else to add? 03:27:46 hongbin: you and sudipto discussed on zun channel, right? 03:28:06 pksingh: i don't remember... 03:28:13 hongbin: :) 03:28:20 Is it related to CPU pinning? 03:29:09 pksingh: if you cann't find the log in zun channel, perhaps, it is a private chat, sorry, i didn't pay attention at that time 03:29:17 Do we support flavor in sun? 03:29:22 hongbin: no problem 03:29:25 lakerzhou: yes, the cpuset feature is about cpu pinning 03:29:49 Zun* 03:30:05 lakerzhou: the native docker driver don't have a flavor, the nova driver will have a flavor 03:30:41 lakerzhou: this bp covers some of the details: https://blueprints.launchpad.net/zun/+spec/auto-select-nova-flavor 03:31:17 Ok, I guess I should read it first 03:31:25 Thank you 03:31:51 ok, if there is no other comment, i am going to advance topic 03:32:06 #topic Discuss BPs that are pending approval 03:32:12 #link https://blueprints.launchpad.net/zun/+spec/support-port-bindings Support container port mapping 03:32:35 this proposal is about expose the port mapping feature in zun api 03:32:54 folks, any thoughts on this one? good idea? bad idea? 03:33:20 we discussed on this one, that this will expose compute node network? 03:33:34 yes, it seems it will 03:34:27 i think we also decided not to do it as of now, if i remmeber 03:34:41 pksingh: ack 03:34:58 others, thoughts? 03:35:15 I think its a good to have feature but not very much urgent for now 03:35:41 shubhams: i think pksingh means this is not a good idea at all 03:36:25 hongbin: I got that and my view point stands same : We could have it 03:36:43 shubhams: i see 03:37:06 shubhams: could you elaborate? why it is good to have it? 03:37:27 pksingh: Can you tell us the reason of not supporting it? 03:37:36 shubhams: would love to consider your input :) 03:37:38 I just want to know the reason behind it 03:37:59 In some (or infact most) scenarios application would want to run on a specific ports and many apps tends to rely on certain ports 03:38:19 mkrai: this needs to go through compute's node network, because port will be of compute node, and we dont want to give access of compute node to anyone 03:38:54 shubhams: this sounds like a k8s service feature 03:39:01 In such scenarios if users of zun could bind ports then it would be easy for them to deploy. They do not need to change anything in their application 03:39:06 hongbin: yes similar 03:39:43 shubhams: ok, i see. i think your requested feature is valid, however, it could be achieved in an alternative way 03:40:06 shubhams: i.e. support k8s service similar feature in zun 03:40:22 pksingh: Do you mean to say, we do not want user to know which ports are free and which are in use ? 03:40:44 pksingh: got your point. But isn't the same way implemented in docker? 03:40:48 shubhams: however, as pksingh said, the port mapping feature has security risk to expose the management network for attack, this is a critial pitfall i think 03:41:03 hongbin: may be that could solve the purpose but I am not sure as of now about it 03:41:13 shubhams: i dont want that user access computenodeip:port 03:41:23 we can have different mechanism to do it 03:41:30 shubhams: ack 03:41:30 hongbin: pksingh : acked 03:42:09 ok, then let me create a bp to record this use case 03:42:23 however, i will leave the solution open, so we could further think about it 03:42:25 mkrai: in docker user is controlling the his node, directly i thnk 03:42:35 hongbin: +1 03:42:59 #action hongbin create a bp to add support for expose container port 03:43:36 any additional comment about this one? 03:44:05 ok, next one 03:44:11 #link https://blueprints.launchpad.net/zun/+spec/support-zun-top Support top command 03:44:30 this proposal is about adding support for 'docker top' command in zun 03:44:38 what do you think abou the idea? 03:45:07 hongbin: How this will work in multihost env ? 03:45:20 +1 for this 03:45:30 same question as shubhams 03:45:44 shubhams: i am not sure exactly 03:45:54 #link https://review.openstack.org/#/c/423971/ 03:46:00 there is patch already for this I guess 03:46:03 shubhams: multihost means? 03:46:19 Is feng present today? 03:46:28 mkrai: i don't think so 03:46:33 pksingh: multiple hosts for container deployment 03:47:05 shubhams: it is getting top logs of the container, i think, what is the challange 03:47:21 IMO, we should have it but we also need a small spec as well 03:47:48 shubhams: +1 03:48:09 shubhams: could you relay your comment to the review? 03:48:24 hongbin: yes sure 03:48:43 shubhams: thx 03:49:13 ok, it looks that is all for this one 03:49:30 #topic Open Discussion 03:49:50 all, any other topic to bring up? 03:50:20 Namrata: are you working on the Zun resource in Heat? 03:50:31 yes I am working on it 03:50:53 Namrata: great, feel free to share if there is any update from your side :) 03:51:33 yeah sure I will add soem patches this week 03:51:38 *some 03:51:42 cool 03:52:57 hongbin: which cycle we are targeting? :) 03:53:18 pksingh: zun is currently at indepedent release mode 03:53:34 pksingh: that means we are free to decide which cycle to target 03:53:42 pksingh: you have any thought about that? 03:53:53 hongbin: ok, cool 03:54:16 hongbin: i think if we can implement all the necesary BPs, then we can decide 03:54:30 pksingh: agree 03:54:54 ok, it looks there is no more topic 03:55:04 all, thanks for joining the meeting 03:55:08 #endmeeting