16:05:21 #startmeeting Large Deployments Team 16:05:22 Meeting started Thu Dec 15 16:05:21 2016 UTC and is due to finish in 60 minutes. The chair is VW. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:05:23 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:05:26 The meeting name has been set to 'large_deployments_team' 16:05:39 #topic roll call 16:05:42 o/ 16:06:04 o 16:06:38 o/ 16:08:05 hey both you guys are here 16:08:06 good 16:08:20 because klindgren has to help me remember things from BCN ;) 16:09:16 Thats a pretty tall ask :-D 16:09:18 with that... 16:09:25 #topic follow up from BCN 16:10:02 * VW goes to dig up some etherpads 16:11:19 here was our etherpad from BCN: 16:11:20 #link https://etherpad.openstack.org/p/BCN-Large-Deployments-Team 16:11:44 I feel like we touched on some things that would be worth getting with Devs and working on forum session submissions 16:13:49 There were some more Neutron bits that came up as potential ones to chase changes on 16:13:56 the client itself 16:14:14 or "partitioning" of neutron - ie, cells 16:14:34 yeah that one in particular stands out to me. although it’s also probably the most vague at this point 16:15:22 there was also some discussion in the Nova meeting about some search/detailed calls in the APIs under cells v2 that left us all scratching our heads a bit 16:15:29 klindgren: you have any thoughts there 16:16:05 Searchlight seems to just re-duplicate cellsv1 16:16:33 i think LDT can be pretty helpful around testing teh cells v1 -> v2 upgrade path, and cells v2 in general. although that’s also a bit of a niche thing. 16:16:49 with a slight change. The fact that they were going to redirect to search light was the most disturbing to me. 16:17:22 I agree mdorman - we should be very helpful in those areas. Sounds like it might be P before we get the upgrade path for v1 users 16:17:29 and yes klindgren - that worries me a lot 16:18:15 I guess, the part I'm trying to figure out is how do we effectively engage in that one? 16:19:34 do we need to have a special meeting and try to get some key nova folks in there? 16:19:43 I think that was wait an see, because they wanted to do what we expected the first time. But the preffered path is searchlight. 16:19:54 that's right 16:20:01 because this is a short cycle 16:20:07 with the shift in dev time lines 16:20:29 but I htink it would be worth raising the concerns about the searchlight path. As it basically duplicates the arch of cellsv1. 16:20:52 but, let me see if I can get john G. or some other folks to come in Feb (since he's UK, it will have to wait two meetings) 16:20:57 so the idea is using searchlight instead of nova calls to do searching of stuff? Or also for listing? I guess i don’t have th e full context on this 16:21:16 yeah that might be best, just some time when we can sync up and make sure we’re all onteh same page 16:21:56 for refference 16:21:58 #link https://www.openstack.org/summit/barcelona-2016/summit-schedule/events/16934/nova-cells-v2-scheduler-searchlight-multi-cell-support 16:22:13 kk 16:23:02 sorry, mdorman - that's the actual session link but the etherpad is in there 16:23:12 yup i made the connection, np :) 16:23:53 are either of you (or anyone else lurking) aware at all of this: 16:23:55 #link https://etherpad.openstack.org/p/ocata-nova-summit-cellsv2-quotas 16:25:18 I was not 16:25:34 i sort of remember skimming over some ML thread about this, but i don’t recall the details 16:27:06 this was another that they wanted our opinion on at the summit 16:27:18 I didn't know if we wanted to craft any "official" feedback 16:28:06 The only official feedback I have is, that quotas in nova is a travesty. So anything that can actually make them not fall out of sync. Would be welcomed 16:31:33 yeah i don’t have any particular thoughts on it either. other than doing quotas in just one place seems like a good way to go. 16:32:06 but it seems like this might get rid of parts of that code path that can and do go wrong. But I dont like the idea of a race causing over allocation. But we are less dependent on that now than we use to be. 16:32:40 is this something we should partner with folks to get forum sessions for in Boston? 16:33:26 looks like the spec already merged, so may or may not be too late for significant feedback. https://review.openstack.org/#/c/388262/ 16:34:19 yeah, although it kind of seems like a subtopic of “cells v2” to me, so maybe it’s most appropriate to aim for a more general cells v2 “cross project” session between LDT and Nova, where we can cover a few related topics. 16:34:58 sounds good 16:35:10 I'll bring this up again next month on the late meeting 16:35:36 and if we can get some Nova folks in in Feb, we should have enough to write a proposal 16:35:43 yup 16:35:57 would time out fairly well for getting stuff prepped for boston as well 16:36:20 yep 16:36:22 I think so 16:36:57 I feel like there was also some glance related stuff we chatted with devs about, but sadly that is escaping me 16:37:22 The other thing I'd like to do is get the us and the new Public Cloud WG to partner on a submission or two 16:37:44 I'll have to follow up with mattjarvis on getting that on his agenda soon for their meetings 16:38:02 VW, good idea ;) 16:39:18 speaking of other groups, is there much overlap with the new “massively distributed” group? from limited info i saw, it wasn’t clear to me how LDT would interact with that one. 16:41:21 yeah, I think there may be some overlap there 16:42:06 I never fully understood that one as much. Massively distributed seems like it just a large collection of smaller clouds? 16:42:16 is that the same as this 16:42:19 #link https://wiki.openstack.org/wiki/LCOO 16:43:17 i’m pretty sure that’s something different 16:44:02 https://wiki.openstack.org/wiki/Massively_Distributed_Clouds i think is it 16:45:15 yeah, we are seeing more and more overalp 16:46:35 side note, we need to clean this up 16:46:38 #link https://wiki.openstack.org/wiki/Large_Deployment_Team 16:47:13 so, I think we need to find out more about the massively distributed group 16:47:18 maybe attend some meetings, etc 16:47:37 Last I heard, the LCCO folks weren't really doing things in IRC, etc, but I'll check with Tom on that 16:47:59 I do think that a good partnership between LDT and the public cloud WG's could produce some great results 16:48:28 mattjarvis: when is the next meeting for Public Clouds? 16:48:30 next week? 16:48:38 next Wednesday 16:48:44 hello all 16:48:49 I just arrived 16:49:01 Massively Distributed Clouds they have this vision of the Fog computing 16:49:17 they want to to do this system where the compute nodes are at the edge near the user 16:49:35 so instead of having a big DC with 1000 Servers, you have 10 servers at each network PoP 16:49:42 ah - thanks zioproto 16:49:47 that is the key idea of Massively Distributed and Fog Computing 16:50:00 I dont know how much overlap there is with Large Deployments 16:50:08 I guess we plan more on the big DC with many machines 16:50:28 so now 16:50:30 for them 16:50:34 practically speaking 16:50:47 the problems are things like running rabbitMQ geographically 16:50:50 So bascially you are more concerned with federation of a bunch of tiny clouds. Trying to act as once large cloud? 16:51:12 that is the vision of the Massively Distributed 16:51:16 kinda like having nova-cells - with a bunch of really small clouds? 16:51:22 you have small amounts of compute nodes in many many sites 16:51:27 s/clouds/cells 16:51:30 I make an examples 16:51:42 if the VMs make NFV things like firewalls 16:51:51 you want these VMs to be spread out and near the end user 16:51:59 its a pretty common telco pattern 16:52:10 mattjarvis: bingo ! :) 16:52:17 I was writing exactly that 16:52:36 yeah, I think there is a TON of overlap between massively distributed and LCCO 16:52:40 so I think we dont necessarly want to follow the same goal of Massive Distributed in our WG 16:52:45 since both are very telco driven 16:53:04 cool 16:53:06 not sure who's in that Massively Distributed group 16:53:15 Oh is the first time I read about LCCO 16:53:20 LCCO ? 16:53:29 LCOO 16:53:34 #link https://wiki.openstack.org/wiki/LCOO 16:53:35 sorry that 16:53:43 * VW types poorly 16:53:45 ;) 16:54:03 anyway, we are about out of time 16:54:07 this was actually very useful 16:54:09 but this is the LDT meeting right ? :) 16:54:12 yep 16:54:19 okay just to make sure :) 16:54:32 we were just talking about the upcoming summit and who we might partner with on some session submissions 16:54:55 sounds like we will engage the nova team (in our Feb meeting probably) to work on some potential topics 16:55:17 mattjarvis: you should probably plan on being here for that if we get some Nova cores in to talk cells V2 16:55:42 and I do think there will be an opportunity to partner with the Public Cloud WG on a submission or two 16:56:04 Neturon client performance/bloat/behavior was another topic that came up a lot with LDT in BCN 16:56:10 sounds good, keep me in the loop - not quite clear yet the direction of travel for Public Cloud WG, but should be defined by then 16:56:30 that we could probably reach out to Neutron on and get some submissions going 16:56:44 sorry - we were all over the place, but this was good 16:56:52 any other things folks need to dicuss? 16:57:22 thanks, mdorman, klindgren, zioproto and mattjarvis for the conversation 16:59:00 sure thing 16:59:22 thank you 16:59:31 bye bye 16:59:46 alright - see you all next month 16:59:51 #endmeeting