22:01:25 <adrian_otto> #startmeeting containers
22:01:26 <openstack> Meeting started Tue Aug 12 22:01:25 2014 UTC and is due to finish in 60 minutes.  The chair is adrian_otto. Information about MeetBot at http://wiki.debian.org/MeetBot.
22:01:27 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
22:01:29 <openstack> The meeting name has been set to 'containers'
22:01:33 <thomasem> o/
22:01:42 <funzo> o/
22:01:42 <sew> o/
22:01:43 <apmelton> o/
22:01:43 <adrian_otto> #topic Roll Call
22:01:47 <apmelton> Andrew Melton
22:02:02 <thomasem> Thomas Maddox
22:02:03 <sew> Steven Wilson
22:02:11 <adrian_otto> Adrian Otto
22:02:11 <funzo> Chris Alfonso
22:02:41 <adrian_otto> #link https://wiki.openstack.org/wiki/Meetings/Containers Our Agenda
22:02:59 <adrian_otto> #topic Announcements
22:03:17 <adrian_otto> first, I am back from vacation, taking up my full load of duties again
22:03:24 <thomasem> Welcome back!
22:03:29 <mtesauro> Matt Tesauro (late to roll call)
22:03:30 <adrian_otto> erw sends his regrets, he will not be able to attend with us today.
22:03:41 <thomasem> Ah
22:03:52 <adrian_otto> I am working on the Containers spec
22:04:13 <adrian_otto> making ascii flowcharts is really time consuming. I was hoping to have a draft for us to discuss today.
22:04:19 <adrian_otto> but soon
22:04:28 <adrian_otto> any other announcements from members of the team?
22:05:35 <adrian_otto> ok, besides the high level spec that I have promised to deliver this week, have any other spec work products been produced?
22:06:04 <adrian_otto> I think we were aiming to make one for the agent, and something for the API for the containers service
22:06:10 <adrian_otto> I could roll that into mine, I suppose
22:06:15 <adrian_otto> the agent will need its own spec
22:06:55 <adrian_otto> did we make any action items last week?
22:07:08 <thomasem> I started a spec for that, let me find a link. Unfortunately, I do not have time given other priorities for my project to work on it at this time.
22:07:11 <thomasem> the agent
22:07:28 <adrian_otto> oh yeah, last week the bot crashed
22:07:33 <thomasem> ah
22:07:35 <adrian_otto> so we don't have minutes from that
22:07:53 <adrian_otto> we have like the first 10-15 minutes or something of logs and that's it
22:08:09 <adrian_otto> thomasem: thanks for staring in
22:08:22 <adrian_otto> maybe we can show what you have so far as a WIP
22:08:53 <thomasem> https://etherpad.openstack.org/p/kilo-container-service-agent
22:08:55 <adrian_otto> and you can revisit it when some time frees up, or another contributor can work on it
22:09:08 <Slower> oh sorry .. o/
22:09:27 <thomasem> adrian_otto ^^ that's what I started with before we wandered off to other topics while in the room.
22:09:31 <adrian_otto> that's a good idea. I can paste what I have so far into an etherpad
22:10:08 <adrian_otto> ok, good, thanks thomasem
22:10:17 <thomasem> Sure thing!
22:10:35 <adrian_otto> #topic Discuss Specs for OpenStack Containers Service
22:10:58 <adrian_otto> so I will go copy what I have into a new etherpad for you to see and provide guidance about
22:11:30 <dguryanov|2> I'm sorry, I've just joined, Adrian, where can I find your spec?
22:12:18 <thomasem> Sounds good to me
22:12:32 <adrian_otto> https://etherpad.openstack.org/p/kilo-containers-service
22:12:47 <adrian_otto> #link https://etherpad.openstack.org/p/kilo-containers-service WIP of Containers Service Spec
22:12:57 <adrian_otto> while I paste more into that you can have a look
22:13:07 <dguryanov|2> Thanks
22:16:20 <adrian_otto> ok, that's all in there
22:16:31 <adrian_otto> only the top couple of sections have my work in therm
22:16:50 <adrian_otto> feel free to edit there if you like
22:18:01 <thomasem> Yes, anyone feel free to pick up that container service agent spec. No idea when I'll get time to work on it.
22:18:35 <adrian_otto> ok, so would anyone like to pledge some time to help advance these documents to the point where we can submit them for review?
22:19:09 <adrian_otto> I'm happy to make a rough cut at each, and erw has indicated his interest as well.
22:19:59 <adrian_otto> I want to be sure that we have a well balanced approach that satisfies a majority of the container use cases for openstack end users
22:20:14 <apmelton> I think until our team meets are internal goals with our current project, it's going to be very hard for us to commit
22:20:29 <Slower> adrian_otto: we're trying to get this in for kilo?
22:20:30 <thomasem> yup
22:20:31 <apmelton> our*
22:20:38 <adrian_otto> Slower: yes
22:20:53 <adrian_otto> apmelton: ok, I understand
22:20:58 <Slower> seems ambitious
22:21:03 <Slower> but may as well try
22:21:18 <adrian_otto> Field of Dreams.
22:21:34 <adrian_otto> if you spec it…
22:21:53 <adrian_otto> anyway I think that once the wider Nova community is looped in, we will not struggle to resource the dev work
22:22:45 <adrian_otto> and I don't mind keeping the team that makes the spec documents small, as that's rather efficient.
22:22:59 <adrian_otto> I do want you all to feel that your input is welcomed and encouraged, and appreciated.
22:23:36 <adrian_otto> so I will keep the following as a standing Agenda item:
22:23:38 <adrian_otto> #topic Status on Proposal Process
22:24:09 <adrian_otto> the current status is that spec documents are WIP, in etherpads, and once reasonably complete will be submitted for review.
22:24:26 <adrian_otto> next week I hope to have links to Gerrit reviews
22:24:54 <adrian_otto> last week we touched on the following topic, and concluded that some work may be needed within Nova…
22:25:01 <adrian_otto> #topic Discuss how to set resource limits (ulimits) on containers
22:25:12 <adrian_otto> I suggest we revisit that dicussion again today.
22:25:16 <adrian_otto> *discussion
22:25:42 <adrian_otto> who can help me restate the problem?
22:26:03 <adrian_otto> ulimits will need to be controlled by cloud operators
22:26:19 <dguryanov|2> in openvz we have user beancounters for that :)
22:26:19 <sew> we need a way to define limits like open files using flavors
22:26:20 <adrian_otto> so having them attached to flavors seems to make sense
22:26:48 <dguryanov|2> I far as I know, you can specify any fields in flavors
22:26:55 <apmelton> extra-specs seem to be the way libvirt has handled cpu quotas
22:27:01 <adrian_otto> once a limit is set, and you are in the shell context of the limited environment, you can't increase the limit
22:27:34 <adrian_otto> so we want to identify options for increasing the limits before the shell context is created
22:27:39 <adrian_otto> did I get that right?
22:28:22 <adrian_otto> thanks apmelton. I'm not yet familiar with extra-specs. Can you explain that to me, please?
22:28:40 <apmelton> adrian_otto: it's basically just metadata for flavors
22:28:45 <apmelton> key/value pairs
22:28:53 <adrian_otto> ok, that's simple enough
22:29:12 <apmelton> sew, do you remember the exact issue that was discussed last week?
22:29:27 <thomasem> It was basically where the topic came from
22:29:30 <apmelton> was it that libvirt didn't natively support it with lxc?
22:29:34 <thomasem> it wasn't supported in libvirt without the patch we now hold
22:29:49 <adrian_otto> when the VM is created, and the Agent started inside it, it is confined by whatever limits were in place when it started.
22:29:50 <thomasem> And we determined that on that front we'll probably want to chat with the libvirt folks about how to get that functionality upstream
22:30:26 <sew> right, we needed to patch libvirt to directly support open file limits based on an optional xml tag in the definition
22:30:55 <thomasem> Furthermore, yep, we discussed using flavors to determine
22:31:27 <thomasem> And so, I think all that remains is handling that extra metadata on the flavor, afaict. Though I haven't looked into what all's involved there.
22:31:44 <thomasem> in terms of getting that working, anyway
22:31:58 <adrian_otto> I think erw raised a concern that there was at least one scheduler filter that may choke in the event that arbitrary metadata is defined on a flavor
22:32:19 <thomasem> Oh, I vaguely remember that. Maybe he could speak more to that next Tuesday??
22:32:24 <adrian_otto> I remember making a remark that we might need to patch any scheduler filters that are limited in that way
22:32:29 <thomasem> ah
22:32:57 <adrian_otto> sure, we can revisit this next time as well
22:33:24 <thomasem> I think in the meantime rcleere was trying to get ahold of danpb to discuss upstreaming the Libvirt functionality.
22:33:30 <thomasem> So, maybe we'll have some update on that front as well.
22:34:16 <adrian_otto> ok, how would you like to list this for a follow up?
22:34:48 <adrian_otto> as an action item, as listed agenda topic? neither?
22:34:48 <thomasem> I'll add an agenda item if we have an update.
22:34:57 <adrian_otto> ok, cool, thanks.
22:34:57 <thomasem> otherwise, no need to discuss it further
22:35:27 <adrian_otto> ok, that brings us to the end of our prepared agenda for today
22:35:31 <adrian_otto> #topic Open DIscussion
22:36:52 <adrian_otto> any further topics to cover, or should we end a bit early today?
22:37:39 <thomasem> Probably end early. Several things in flight; should have more to talk about next Tuesday along with more attendance.
22:37:41 <adrian_otto> anyone else who would like to be recorded in attendance today, this is the time to chime in with a wave
22:38:11 <Slower> I guess no work has begun on the new container service?
22:38:15 <Slower> no repos yet?
22:38:26 <adrian_otto> Slower: correct, not yet
22:38:28 <Slower> ok
22:38:41 <adrian_otto> Michael Still reached out to me about this by email
22:38:56 <adrian_otto> and we will work together this week to sort out how best to start this
22:39:28 <adrian_otto> at the midcycle we talked about using Stackforge, but it's possible we might just put another repo in under the compute program
22:39:49 <adrian_otto> and skip that step, much like we did with Cielometer.
22:40:04 <adrian_otto> did any of you have a preference?
22:40:46 <adrian_otto> we will also want to select an initial core team, so if you have an interest in that, please contact me about it
22:41:22 <Slower> are we going to have a meetup in france?
22:41:23 <adrian_otto> ideally we get a mix of subject matter experts, and nova cores
22:41:49 <adrian_otto> Slower: yes, some of the design summit agenda for Nova will be allocated for this.
22:42:37 <adrian_otto> any suggestions for what to put into a few session topics?
22:42:59 <adrian_otto> I thought it might be good to re-propose the high level concept like we did at the Midcycle
22:43:13 <adrian_otto> so that a wider audience can grok what's coming
22:43:44 <Slower> that sounds good to me
22:43:47 <adrian_otto> another one for discussing the design of the agent
22:43:53 <Slower> especially since I missed the midcycle
22:44:28 <thomasem> transport mechanism design
22:45:03 <adrian_otto> Slower, did you look at this:
22:45:22 <adrian_otto> #link https://wiki.openstack.org/w/images/5/51/Containers_Proposal.pdf Containers Proposal from Nova Midcycle Meetup
22:45:46 <Slower> adrian_otto: I did thanks
22:45:57 <adrian_otto> thomasem: yes, I'll encourage you to propose that as a summit session
22:46:17 <adrian_otto> as soon as that opens. That usually happens about a month before the summit
22:46:22 <adrian_otto> so probably in a few more weeks
22:46:47 <adrian_otto> ok, anything else before we wrap up?
22:46:59 <sew> guess i'd like to see the design goals restated a bit more prominently in whatever proposal is given in france
22:47:31 <adrian_otto> sew, that's a great suggestion. I'll do my best on that for you.
22:48:01 <sew> thx adrian_otto
22:48:36 <adrian_otto> ok, see you all next Tue Aug 19 at 1600 UTC
22:48:44 <thomasem> yep yep, take it easy!
22:48:51 <adrian_otto> thanks!
22:48:52 <adrian_otto> #endmeeting