21:01:33 #startmeeting quotas-wg 21:01:33 Meeting started Mon Feb 29 21:01:33 2016 UTC and is due to finish in 60 minutes. The chair is nikhil. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:01:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:01:37 The meeting name has been set to 'quotas_wg' 21:01:42 hi 21:01:46 hey there 21:01:49 #topic roll call 21:01:59 vilobhmm 21:02:02 o/ 21:02:07 I see vilobhmm11, ninag and mc_nair already here 21:02:20 Let's wait a couple mins for folks to show up 21:02:58 sure 21:03:06 Let me highlight reminder 21:03:08 Courtesy meeting reminder: nikhil, vilobhmm, DuncanT, mc_nair, 21:04:04 lets start 21:04:17 #topic agenda 21:04:20 Hi. Posting from my mobile, so a touch lagged, sorry 21:04:25 #link https://etherpad.openstack.org/p/quotas-wg-meeting-agenda 21:04:29 HI DuncanT 21:04:35 thanks for adding items to the agenda 21:04:42 great to see input already! 21:04:42 +1 21:05:00 This is our first meeting 21:05:26 so the purpose is for everyone to get introduced, express level ofinterest and define scope and objectives 21:05:49 We also wanted to get some early eyes on the spec and establish momentum in the freeze period 21:06:11 so that we can have some clear, concise and productive conversations at the summit 21:06:48 Also, one aim is to gather input at this close-nit gathering and then share insights to the wider audience from time to time at the regular cross project meetings 21:07:14 So, let's start with a very brief introductions, interests etc. 21:07:21 #topic Introductions, interests, objectives, etc. 21:07:52 I can go first in people prefer.. 21:08:17 sure :) 21:08:39 I am Nikhil Komawar, primarily focusing on Glance and Searchlight :) 21:09:10 I am interested in this work to get cross project input to glance for the quotas work in the coming up cycle and help prioritize it. 21:09:35 hey - I'm Ryan McNair - work on Cinder and have been doing a lot of work with Nested Quotas in Cinder for the past month or so 21:09:44 I'm DuncanT, cinder core, and my main interest is in seeing this attempt not repeat the mistakes that the cinder nested quota code made, which can be subsided quite easily in one line :-) It should be noted in terrible at actually finding time to write code, but I'm pretty good on reviews 21:09:58 Duncan Thomas, HPE currently 21:10:12 ++ 21:10:20 vilobhmm11: you're next 21:10:31 vilobhmm11: I am aware that you 21:10:37 you've a conflicting mtg 21:10:44 so if you want me to move please send a ack 21:10:50 * smcginnis is lurking 21:10:59 vilobhmm I am Magnum core. I have worked in the areas of Quotas in Nova/Cinder/Magnum. Problems with quotas in normal use-case and at scale. 21:10:59 :) 21:11:03 First to introduce and implement Nested Quota in Cinder 21:11:12 Working on Quota Flavor Classes, Quota by Flavor, Quota by Availability Zone, Quota by X 21:11:29 so have worked/focussed more on quota in general :) 21:11:44 the aim is to make a common entity 21:11:53 which will be useful for all the projects 21:12:10 and new upcoming projects who plan to use quota does nto have to re-invent the wheel 21:12:17 and spend additional cycles 21:12:50 ’m Nina Goradia…interested in getting hierarchical quota support in Glance..would like it to be consistent across projects, but am open to whether or not it needs to be a common code base 21:13:47 vilobhmm11: thanks, glad to get your insights! 21:13:53 ninag: and that sounds like a good point to discuss early on! 21:14:35 I guess those are all for today. 21:14:47 Thanks everyone, glad to know the interest groups. 21:14:56 nikhil : Few points we should dicsuss early on : #1. whether common code base (Y/N) #2. If common code base then should it be a library or a service 21:15:23 just to add to ninag 21:15:26 vilobhmm11: sounds good, should we do that for open discussion today and carry it on next week? 21:15:56 nikhil : open discussion sounds good to me; 21:16:00 cool 21:16:08 Moving on 21:16:14 #topic Spec (vilobhmm) 21:16:22 #link https://review.openstack.org/#/c/284454/ 21:16:32 vilobhmm11: mind giving us a tl;dr; version? 21:16:43 sure 21:16:46 (for record keeping here in the logs) 21:16:58 so would request everyone to have a look and give there feedback 21:17:44 right now i don't want to get into library vs service debate as that is the main reason why this *idea of common quota thingee* didn't made in the past 21:17:50 so calling it an entity 21:17:57 we can discuss what it should be 21:18:00 library/service 21:18:51 so goal here is to have a common *entity* which #1. minimizes or zeros down the code duplication #.2 New project and use it and enable quota #3. New features like nested quota, quota by x can be added on demand 21:19:16 #2.New project can use it and enable quota 21:19:48 that is the *WHY* part 21:20:09 now about *HOW* to do it have included details in the "Proposed change" section please have look 21:20:24 all feedbacks are welcome :) 21:20:47 ++ 21:21:10 I guess that sort of pushes the discussion around the scope of the cross-prj initiative 21:21:19 cool, will take a look at the spec right after the meeting 21:21:20 yup 21:21:35 To me it seems like we really need to understand the use cases for proposing a new project 21:21:38 hence renamed the spec as well to "Delimiter - Cross Project Quota Enforcement and Management" 21:21:58 nikhil : sure 21:22:14 It would be a overhead for maintaining a new service for quotas if projects can't implement it internally 21:22:34 And that's another part of the scope discussion comes into play 21:22:51 I would like to know 21:23:10 nikhil : Regarding use-case for a new project/library/entiry - few details are covered here https://review.openstack.org/#/c/284454/4/specs/quota-library.rst as part of "Problem Description" 21:23:53 vilobhmm11: ++ 21:24:08 nikhil : three of the most common problems that were seen as part of ops meetup and previous dev meetup are jotted down here https://review.openstack.org/#/c/284454/4/specs/quota-library.rst 21:24:12 I agree on the technical implications and the optimizations aspects of things 21:24:48 I would agree to those 21:25:13 and have had the pleasure of experiencing the #3 of that many a times first hand ;) 21:25:32 my use case concerns were around: 21:25:34 1. Are there custom glitches in projects around quotas? (one feature/resource dependent on another in subtle way) 21:25:48 I can't follow all those links right now, sorry, I'll have to catch up later 21:26:29 2. Do projects need to have DB migrations that would affect quotas work and that can't be handled outside of the project scope? (May be upgrades, or quota lib adding a dependency to oslo.vo?) 21:26:50 DuncanT: np 21:27:04 these are good points and i think we should cover it as part of spec 21:27:09 nikhil : ^^ 21:27:19 nikhil: for 1) are you referring to quotas that are related/dependant? 21:27:20 i can update the spec with more details 21:27:41 3. If done as a separate project, how do we take care of cross-resource transactions within the project? 21:28:04 mc_nair: yes, quotas that are dependent 21:29:06 vilobhmm11: the #3 is my perspective of images and image members in glance. 21:29:27 members are not independent of images but 21:29:45 image <-> members have a many to many relationship 21:30:19 oops, I realized that we're running over. 21:30:34 For some reason, forgot this was a 30 mins meeting. sorry about that. 21:30:43 Do folks mind carrying this for another 5 mins? 21:30:56 I do not think there's another meeting in this room. 21:30:59 I don't mind 21:31:04 ok great. 21:31:23 anyone else, have objections? 21:31:49 nope, I can continue 21:32:06 I'm fine continuing 21:32:12 #info meeting to be carried on till 2135 UTC 21:32:39 quickly running out of time on that too :) 21:32:40 So, to better define the scope , I think we need a wider use case feedback. 21:33:01 And thanks vilobhmm11 for willing to add some initial thoughts on those in the spec. 21:33:15 It would be good to get them on the review or during CP meetings. 21:33:27 nikhil : sure…but this problems is only for existing projects who have quotas right now..and we need to think on migration plan before the switch happens 21:33:36 Just to get my 5 cent worth in, my main point was going to be that you need to specify behaviours, including corner cases, in extreme detail, from the beginning... Figuring then out as you go doesn't work out, and writing the tests first is probably a good way to proceed 21:34:19 vilobhmm11: and DuncanT: both are great points indeed! 21:34:30 but they conflict a bit ;) 21:34:49 I would like us to think about the migration and if that's a possibility.. 21:34:56 but the first point to answer is: 21:35:04 1. separate repo or not 21:35:11 2. if yes, lib or project 21:35:31 nikhil : those are the starting points imho 21:35:34 and that would be essential to get from other projects too I think 21:35:51 and we're out of time again :) 21:36:04 Separate repo is the only way that I can see this making progress without falling into the mess of fixing a specific project's bugs 21:36:22 Lib or project I don't care 21:36:22 +1 21:36:35 ok, that sounds good. 21:36:43 if we have a concensus on 1. separate repo or not 2. if yes, lib or project we have a clear path and should then think of migration, testing 21:37:07 what do ppl think for adding a voting section on the spec on this topic? 21:37:17 nikhil : ^^ 21:37:20 this == separate repo or not 21:37:23 What is the practical difference between a lib and a project? 21:37:42 DuncanT : library vs service (project) 21:37:45 DuncanT: project would mean a API backed by DB 21:37:47 is what nikhil meant 21:38:16 Got you. Thanks 21:38:19 more like: Quotas as a Service (QaaS) 21:38:57 The downsides of going through an API are a) failure modes and b) latency 21:39:08 I think a separate repo makes sense..and if the behavior is fleshed out and documented..then it will help decide whether it meets all needs from projetcs 21:39:19 The upsides are debugability and consistency 21:39:34 or whether we have the cross-resource dependencies..and will help towards the library or project decision 21:40:12 I think the consumer entry points are pretty much the same whether we go project or lib (reserve, commit, unreserved, list, edit, etc) 21:40:20 DuncanT: yeah, I think another downside is "concurrency issue" for intra-prj resources 21:41:00 DuncanT: yes, agreed on the entry points. 21:41:31 ok, I think we'd call off on the meeting today. We can definitely continue the discussion on #openstack-dev if we like. 21:42:28 I think we've got plenty to think about anyway 21:42:34 sorry for running over and we will try to be more precise from next week onwards in case people have conflicts and prior commitments. 21:42:46 sounds good 21:42:49 DuncanT: agreed 21:42:52 Thanks all for joining. 21:43:07 #endmeeting