16:00:39 #startmeeting containers 16:00:40 Meeting started Tue Feb 3 16:00:39 2015 UTC and is due to finish in 60 minutes. The chair is adrian_otto. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:41 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:00:43 The meeting name has been set to 'containers' 16:00:45 #link https://wiki.openstack.org/wiki/Meetings/Containers#Agenda_for_2015-02-03_2200_UTC Our Agenda 16:00:49 #topic Roll Call 16:00:57 o/ 16:00:57 Adrian Otto 16:01:00 Andrew Melton 16:01:00 Steven Wilson 16:01:06 Hongbin Lu 16:01:39 o/ 16:01:39 ward harold 16:02:38 hello apmelton, sew, hongbin, prad, and wkharold 16:02:48 we will proceed in just a moment 16:02:48 Digambar Patil 16:03:24 hi 16:04:05 hello diga and tartinette 16:04:23 Hi Adrian 16:06:50 ok, welcome everyone. Let's begin! 16:06:53 #topic Announcements 16:07:07 1) Welcome Hongbin to magnum-core 16:07:25 Thanks. It is my hornor to join the great team 16:07:28 congrats hongbin 16:07:37 Hongbin has been a consistent contributor and reviewer for Magnum, and has been recognized as a key member of our team 16:08:01 we trust Hungbin to help us with core reviewer responsibility 16:08:02 congrats hongbin 16:08:14 thx 16:08:37 so we are thrilled to add you, thanks! 16:08:42 2) Reminder, next planned release date is 2015-02-16. 16:09:41 if we have any team members needing guidance to determine where their efforts may be best focused to help us reach this milestone, please chime in during Open Discussion today, in #openstack-containers at any time, or feel free to contact me individually. 16:10:06 3) Reminder: adrian_otto will be away on 2015-02-17 for our 1600 UTC meeting. Our pro-tem chair will be sdake. 16:10:20 are there any other announcements from team members? 16:11:00 ok, proceeding... 16:11:03 #topic Review Action Items 16:11:12 (1) adrian_otto to break https://blueprints.launchpad.net/magnum/+spec/magnum-docker-backend-selection into multiple blueprints 16:11:23 thanks to help from apmelton, this is complete. 16:11:37 are there any questions about the breakdown of this work? 16:12:10 adrian_otto: I suppose we'll discuss which milestones for each piece during task review? 16:12:21 Yes, sir. 16:12:25 alrighty 16:12:34 ok, so we will revisit this one in a moment. 16:12:39 (2) adrian_otto to follow up with yuanying about ironic-heat-template 16:12:54 this is pending, so I will carry it forward 16:13:01 #action adrian_otto to follow up with yuanying about ironic-heat-template 16:13:11 (3) adrian_otto open a blueprint on docs for things to write for use with M2 release 16:13:29 #link https://blueprints.launchpad.net/magnum/+spec/m2-docs Documentation for M2 16:13:41 this is where we will track our documentation work items. 16:14:04 If you are able to contribute to any documentation efforts, let's be sure to mention them in this BP. 16:14:26 (4) adrian_otto to open a bug/task ticket for cherry-pick of updates in the larsks heat template code. 16:14:38 #link https://bugs.launchpad.net/magnum/+bug/1417642 Pull in updates from larsks heat template repo 16:14:53 (5) adrian_otto to add ceilometer integration as a mid-cycle discussion topic. 16:15:12 #link https://etherpad.openstack.org/p/magnum-midcycle-topics Midcycle Discussion Topics 16:15:23 feel free to add to the topic list. 16:15:42 that concludes our action item review 16:15:57 #topic Mid-Cycle Meetup Planning 16:16:04 #link http://lists.openstack.org/pipermail/openstack-dev/2015-January/054053.html Call for Feedback 16:16:23 so, I proposed new dates for the Midcycle that land in the following week 16:16:39 please take a moment to respond to the Doodle Poll 16:16:53 #link http://doodle.com/ddgsptuex5u3394m Select a date 16:17:02 we are looking for two consecutive days 16:17:46 I think we need to take a decision on this BP - https://blueprints.launchpad.net/magnum/+spec/magnum-scheduler-for-docker 16:17:49 the Feb 23+24 and 25+26 are too contended for scheduling, so I proposed Mar 2+3 16:18:07 diga: We will address that in task review, one moment please. 16:18:20 ok 16:18:58 ok, so I look forward to your responses to the Mar 2-3 dates on the Doodle poll. 16:19:03 #topic Blueprint/Task Review 16:19:18 so, first I will call on apmelton, and second on diga 16:19:35 are there other team members who would like to discuss a work item? 16:20:15 ok, apmelton. Please begin on https://blueprints.launchpad.net/magnum/+spec/magnum-docker-backend-selection 16:20:57 at this point, I think https://blueprints.launchpad.net/magnum/+spec/docker-bay-heat-template is the only one I'd feel comfortable pulling in for M2 16:21:14 ok, let's take a look at that one 16:21:15 heat templates are pretty new to me, so there will be a bit of a learning curve 16:21:29 ok, so definitely touch base with sdake 16:21:39 he's a Heat expert, and can lend a hand with that 16:21:47 awesome, will do 16:22:12 you might also consider Randall Burt 16:22:32 he;s arguably one of the more talented engineers who have worked on heat and templates 16:23:02 rephrased. He is definitely talented, and very experienced in this subject matter. 16:23:27 LMK if you want additional Stackers who can help. 16:23:46 ok, I'll let ya know 16:24:14 I have set that blueprint as direction-approved 16:24:24 and set to m2 as the series goal. 16:24:47 anything more on this topic? 16:25:02 don't think so 16:25:04 o/ 16:25:18 apmelton: will you be claiming that as the Assignee? It's currently set to None 16:25:26 hello thomasem 16:25:37 adrian_otto: sure, just assigned myself 16:25:57 ok, diga, you are up now. Let's discuss scheduling as a topic. 16:26:18 this is something we will also open for discussion on the ML to get wider input from the Nova/Gantt team 16:26:20 yep 16:26:50 yes 16:27:42 #link https://blueprints.launchpad.net/magnum/+spec/magnum-scheduler-for-docker Add schedulre for native docker support 16:29:03 our initial thinking on this topic was that the first scheduler (Gantt alternative) would allow the caller to specify a node to create a container on, or specify a node to not create a container on. 16:29:25 so, one thing to consider is that swarm does offer scheduling 16:29:28 we would implement a simple (pluggable) scheduling algorithm for sequential-fill 16:29:52 apmelton: yes, it has a naive scheduler 16:30:05 ditto for k8s, yes? 16:30:13 wkharold: yes, that's right 16:30:32 adrian_otto: when was the last time you checked out swarms scheduling? 16:30:39 ideally what we build (or integrate) for native docker would later be used for other container backeneds 16:30:46 browsing through the repo, it appears to actually have quite a few features 16:31:07 https://github.com/docker/swarm/tree/master/scheduler/filter 16:31:15 the key issue is how to manage the placement of containers within a pod on different nodes, based on some user-selected logic 16:31:52 oh, sweet, this is under active development again! 16:32:05 I am going to link that in the PB whiteboard… one sec 16:32:16 can I get some code from nova/cinder scheduler for initial implementation ? 16:32:16 s/PB/Blueprint/ 16:33:41 diga: I think we should consider that as an option, yes 16:33:58 okay 16:34:02 we should view any work we do on scheduling in the short term as potentially disposable 16:34:26 so we don't need to be too picky about it 16:34:43 okay 16:35:22 one potential drawback of using Swarm is that we need to load and run the Swarmd binary (and associated agent), and place them properly. 16:35:49 whereas, if we used code from nova, it may be possible to leverage nova-compute as an agent 16:36:10 thoughts on this? 16:36:19 for swarm, is that possible to plug a custom filter? 16:36:44 we'll need to look at that code to find out 16:36:48 yes, using nova sch code is great advantage for us 16:37:04 k 16:37:54 we did get some feedback from the Nova team on the status of Gantt 16:38:19 they indicated that we should not expect movement on Gantt during the kilo release. 16:38:53 I am generally reluctant to suggest that we pick up development of Gantt because it really needs a bunch of high-risk code to be merged into Nova 16:39:05 hi 16:39:12 so that effort should really be driven by the Nova team who can make that happen. 16:39:34 yes 16:39:41 I worry that if we begin proposing code for Nova that's not a priority for the Nova team, that it will rot in perpetual review 16:39:49 and just be a waste of time on both ends 16:40:07 hongbin: a glance at the swarm filter code shows just four builtin filters, go doesn't support dynamic code loading so customer filters seem a long shot 16:40:19 so it's probably smarter to engage on that once it ascends in priority 16:40:43 s/customer/custom/g 16:40:55 wkharold: thx. I see 16:41:08 wkharold: thanks for checking on that 16:41:13 np 16:41:16 custom filters might not be that important 16:41:31 so let's be sure not to rule out an option because of that 16:41:49 I think we can all agree that it's definitely nice-to-have, but I'm not convinced that feature is essential 16:42:09 hi 16:42:09 make sense 16:42:12 if we have a filter scheduler that's reasonably complete, it's likely to be useful for most needs 16:42:37 tartinette: you are welcome to add in any time if you want 16:43:24 adrian_otto: I wonder how much we could get by with the constraint filter,, basically custom labels are applied to the docker daemon running, and then the user can filter based on those labels 16:43:44 #action adrian_otto to initiate an ML thread for discussion of native Docker scheduling support for Magnum 16:44:20 apmelton: I think that would work. Do you think that would be elegant from an end-user perspective? 16:44:48 seems pretty simple… add labels, then restrict on them. 16:45:10 yup, of course you couldn't have very sophicated filtering logic though 16:45:55 is that similar to something like affinity? 16:46:01 We could put a hook in (either in swarm or something of our creation) to allow for an external filter 16:46:18 hongbin: affinity is a separate filter in swarm 16:46:23 the API definition for that would likely be pretty simple 16:46:41 apmelton: I see 16:47:52 ok, I think this was a good start. We will get this discussion going ont he ML as well to poll input from other subject matter experts. 16:48:07 I took that as an action item earlier. 16:48:30 okay 16:48:42 any final thoughts on this before opening discussion to other work items? 16:49:06 any other work items that we should discuss today as a team? 16:49:27 ok, then I will open us up to all topics now… 16:49:35 #topic Open Discussion 16:50:55 anyone had issues being able to pull docker images after doing a devstack deploy per the quickstart? (feel free to tell me to take it offline is this is the wrong forum) 16:51:44 wkharold: your question is appropriate here 16:52:01 wkharold: docker images being pulled from the atomic nodes? 16:52:21 being pulled to the atomic nodes 16:52:40 wkharold: I encourage you to document your observations in a but ticket here: https://bugs.launchpad.net/magnum/+filebug 16:52:47 s/but/bug/ 16:53:08 k, trying to reproduce this AM 16:53:14 There may be something we can learn from it to improve the quickstart instructions 16:53:23 I think it would be great for Magnum to provide API for image/snapshot management 16:54:03 hongbin: are you thinking of something generic, or something Docker backend specific? 16:54:13 generic 16:54:24 like nova image-list 16:54:35 do the glance people have any designs on managing container images? 16:54:51 wkharold: we have not explored that yet 16:55:07 that's something I will add to the midcycle as a potential design topic 16:55:08 seems like the ultimate place for it 16:55:30 hongbin: I think nova image-list was created before Glance existed 16:55:48 and it's probably still there for vestigial reasons 16:55:59 wkharold: agreed 16:56:30 adrian_otto: k. my thinking is 16:56:51 adrian_otto: not related question, is there any way to join online to midcycle meetup ? 16:56:51 in long term, we needs an abstraction to manage images for different container backend 16:57:28 like Rocket? 16:57:48 not sure what is Rocket 16:58:11 Docker alternative (sort of) 16:58:33 wkharold: I see 16:58:42 hongbin: CoreOS proposed a container format, and offered Rocket as a reference implementation 16:59:05 https://github.com/coreos/rocket 16:59:36 ok, time for us to wind down for today 17:00:22 Thanks to everyone for attending. Our next meeting is 2015-02-10 at 2200 UTC. I look forward to seeing you then! 17:00:25 #endmeeting