13:59:43 #startmeeting Glance 13:59:44 Meeting started Thu Dec 3 13:59:43 2015 UTC and is due to finish in 60 minutes. The chair is flaper87. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:59:45 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:59:48 The meeting name has been set to 'glance' 13:59:55 #topic Roll Call 13:59:57 o/ 13:59:59 o/ 13:59:59 o/ 14:00:07 o/ 14:00:31 mfedosin will be here in a minute 14:00:31 * flaper87 has a quite flaky internet connection... expect some lags 14:00:38 ativelkov: thanks 14:00:40 o/ 14:00:49 o/ 14:00:51 o/ 14:00:55 sabari: O.O 14:01:01 sabari: glad to have you here 14:01:02 o/ 14:01:04 :D 14:01:06 o/ 14:01:09 * flaper87 hands a cup of coffee to sabari 14:01:18 #topic Agenda 14:01:21 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:01:25 flaper87 The jetlag helped a bit :D 14:01:26 That's our agenda for today 14:01:31 o/ 14:01:46 o/ 14:01:47 as usual, feel free to put stuff in there if you would like to talk about them 14:02:18 One thing w.r.t the agenda, now that we're talking about it. If you need to raise awareness on a specific spec, you should feel free to add it to the drivers agenda 14:02:36 It's fine and perfect to discuss it here if you need a broader community to chime in 14:02:52 but, if it's just to raise awareness and ask for review, remember we meet on Tuesdays to talk about that 14:03:00 flaper87, ok, got it 14:03:05 you're not required to attend the meeting, we'll review the spec even if you're not there 14:03:08 :) 14:03:16 #link https://etherpad.openstack.org/p/glance-drivers-meeting-agenda 14:03:26 that's the link to the driver's agenda 14:03:30 you're spoiling the cats 14:03:30 coolio, moving on 14:03:38 :D 14:03:41 #topic Updates 14:03:52 Artifacts: 14:03:55 ativelkov: mfedosin ? 14:04:09 we met for a mins 14:04:13 the bot wasn't there 14:04:20 had a convo for about 20 mins 14:04:26 nikhil: :( 14:04:32 the gist of it is that we are waiting for the spec 14:04:42 I am back to active work (sorry, took so long, but finally got the approve). We plan to have a Glance v3->Glarev0.1 spec to be ready by the next drivers meeting 14:04:45 to be written or reviewed ? 14:04:45 and some hint of thinking from ativelkov 14:04:58 ativelkov: awesome 14:05:05 And so will the refactored API spec, hopefully 14:05:06 written, but I will let ativelkov chat 14:05:10 That sounds great! 14:05:35 * flaper87 is looking forward to that spec 14:05:46 and as I promised I will prepare a spec about moving v3 to standalone service 14:05:50 to be written, yes. We've also got some commitment from mfedosin and dshakhray to do the actual migration of code 14:06:14 nice, sweet! 14:06:21 Thanks folks 14:06:24 ativelkov: so to clarify, we are porting and not delete then write again ? 14:06:40 for now - porting, yes 14:06:41 I think I -2'd the deletion patch 14:06:47 as I think porting is the way to go 14:06:49 cool 14:06:54 thanks, deletion sounds risky 14:06:56 Just moving the current v3 to glare v0.1 14:07:00 ok, drivers updates 14:07:04 I abandoned this patch 14:07:06 experimental prolly doesn't give us that contract right 14:07:22 thanks 14:07:29 Notm uch from our side. There have been more discussions on the new image import process and an attempt to simplify it 14:07:48 A new PS will be proposed and I hope we'll reach an agreement before the next drivers meeting 14:08:07 NExt week we'll start making decisions based on what we have because we need to move that forward 14:08:08 I have solid feedback and 2-3 use cases that will help shape it well 14:08:19 other than that, the db purge spec landed 14:08:26 and more reviews to the rest of the specs are coming 14:08:27 I have already communicated that to Brian and writing it down in a formatted manner 14:08:36 nikhil: awesome :D 14:08:39 thanks 14:08:42 rosmaita: ^ 14:08:47 moving on 14:08:54 Updates on the Nova v1 -> V2 work 14:09:07 I have good news 14:09:09 One of the feedbacks I got was that we hadn't provided news on what's going on there 14:09:18 Which is true 14:09:22 so, here's some feedback 14:09:26 actually, mfedosin you go for it 14:09:41 currently we're able to do listing with v2 14:09:52 w0000h0000000 14:09:53 nice! 14:10:01 changes-since works perfect with new filters 14:10:01 ship it! 14:10:08 great! 14:10:09 some patches have been merged already 14:10:14 nnnnice 14:10:14 nice nice nice 14:10:18 and all tempest tests pass 14:10:20 awesome! 14:10:26 well done! 14:10:37 delete and download work as well 14:10:41 cool! 14:10:47 The nova core team is falling behind with some reviews on this work but I'm working with them to make sure there's nothing blocking reviews 14:10:59 I need to concentrate and finish with 'update' today 14:11:06 There's a series of patches up for review: https://review.openstack.org/#/c/231696/ 14:11:26 That is, things have been implemented and we mostly need to land them 14:11:46 mfedosin: thanks for the updates 14:11:53 #topic Cross project: backwards compat for libraries and clients (nikhil) 14:11:54 welcome :) 14:11:55 nikhil: floor is yours 14:11:57 mfedosin et.al : excellent work 14:12:12 on the x-prj topic 14:12:30 now that we will be (semi-)enforced to implement this 14:12:44 (I will let the spec talk the details) 14:12:58 what do we all think about glance-store api refactor? 14:13:01 #link https://review.openstack.org/#/c/226157 14:13:16 I am not sure how feasible it is to keep that 100% back-compat!? 14:13:30 thanks flaper87 14:13:49 or do we mark this lib as an exception 14:13:59 I still think we should do it. The whole point about this refactor was not to delete everything and re-implement. The idea is to write the new API alongside the old one 14:14:11 so that it'll give time to consumers of the library to migrate 14:14:26 I think our current API is pure crap (pardon my language) 14:14:30 yeah, but then we need to maintain 2 APIs 14:14:37 and it was thought for internal use in Glance 14:14:44 nikhil: sure but we'll deprecate the old one 14:14:53 it's just to give time to projects to move to the new one 14:15:04 In openstack, glance is the only one using it for now 14:15:06 which makes it easier 14:15:08 so you are saying that store would need an exception 14:15:30 I don't think that spec says we can't deprecate stuff 14:15:43 What it says is that we *have* to commit to deprecate stuff 14:16:18 is this a priority for mitaka? 14:16:18 The plan with that refactor has always been to deprecate the old API and give time to migrate 14:16:27 it is not 14:16:39 but I believe nikhil is referring to the spec 14:16:46 which needs lots of work 14:16:58 I think it'll take a couple of months for that spec to be ready 14:17:07 which means it'll fall into the N cycle 14:17:20 flaper87: the back compat one or the refactor one? 14:17:43 nikhil: the refactor one 14:17:52 ok 14:18:13 but even with back compat one it's a painfully and hard to commit process 14:18:35 for instance supporting 2 APIs for 4 cycles is literally saying that you never get rid of it 14:18:45 the amount of catch up needed is humoungous 14:18:50 We can refactor the refactor spec and do a more progressive work towards something saner 14:19:50 even progressive work 14:19:54 I hate to assume stuff but here I go: I'd assume that many issues since migrating glance to the new glance_store API is most of the work 14:20:09 doesn't guarantee less catch required 14:20:23 True but at least it makes it more manageable 14:20:36 nikhil: may I ask what your suggestion is? 14:20:43 sure the API is, but not the x-prj enforcement 14:20:56 I think we have 3 choices: 14:21:08 1. let things be as is and progress with x-prj requirements 14:21:43 2. add exception to x-rpj list of libraries that do not warrant a API back compat contract, so that we can port it in a cycle or max 2 14:22:30 3. put back store in glance or mark it as private library (or one w/o API guidelines) (we can refactor the API if it remains married to glance) 14:22:40 I personally think we either need a new api, or to move it back into the glance tree 14:22:56 so, to me, #3 is out of discussion 14:23:00 currently the library is a real pain to use 14:23:10 As of now, I'd go with #2 and add the exception 14:23:36 do we need an exception for glance_store ? The spec seems to say we need to adhere to semver and can make breaking changes, right ? 14:23:42 Fix the API, make it consumable by other projects that still want to consume it (a.k.a Nova) and then commit to the backwards compat policy 14:23:48 sure, but still sanely maintainable. I fear a large refactor could potentially disrput any perf/reliability enhancement done 14:24:19 That we need to discuss on the spec. I think the whole way the spec started gave the wrong impression of what the goal is 14:24:40 A fourth option would be to start a whole new library and let glance_store die 14:24:57 That new library would do what glance_store is supposed to do but w/ a new API 14:25:14 sabari: seems to have a hidden clause. the feeling I got was breaking directly wasn't allowed 14:25:40 ok, we need to move on so, to have some action items 14:25:42 flaper87: sure, #4 is a good option too 14:25:51 Can we mention this on the x-proj spec? 14:25:56 I will 14:26:03 wanted to get early glance feedback, thanks/ 14:26:08 We need a way to be able to either refactor glance_store or create a new one 14:26:22 I'm sure good feedback will come out from that spec and ppl following it 14:26:32 nikhil: thanks for bringing this up 14:26:50 #action nikhil to bring the glance_Store refactor discussion on the x-proj compat spec 14:27:02 ok, moving on unless there's something else critical here 14:27:18 #topic Cross-project work (flaper87) 14:27:36 on a similar note to what nikhil just mentioned, I think we need to improve our x-proj presence 14:27:47 nikhil: I believe you're the x-proj liaison, right? 14:27:55 yes 14:28:01 coolio 14:28:12 what do you think about having x-proj updates at the beginning of our meeting ? 14:28:19 Just like artifacts and drivers 14:28:31 it's mostly formalizing the process and getting some chasers to put on the list of people 14:28:40 flaper87: surely, I think the p 14:28:49 though the proposal was 14:29:07 you need to have a comment on the x-prj spec 14:29:26 but prolly still developing 14:29:40 yeah 14:29:52 I can request feedback in the meeting or outside depending on time constraints whatever you all prefere 14:30:03 I'm currently worried about our presence in the x-proj meetings and how much feedback we're giving on that area 14:30:15 I'm wondering if we should have 2 liaisons so that you have some help there 14:30:21 heh 14:30:28 we had one x-prj meeting in 3 weeks 14:30:35 yup 14:30:55 ok, lets start by having an x-proj specific topic in our meetings 14:31:01 I think that'd help a lot 14:31:09 ok 14:31:29 nikhil: btw, don't get me wrong. I think you're doing a great job. I hope this topic didn't come out the wrong way 14:31:36 and I am just back settling in catching up with backlog :/ 14:31:47 flaper87: yeah, np 14:32:15 coolio, moving on unless there's something else here 14:32:29 #topic Community Recognitions 14:33:28 So, I don't really have a way to recognize ppl's work other than mentioning it publicly in our meetings. I think some folks have been doing an amazing job and I would like to, at the very least, recognize that here in a per-milestone basis 14:33:49 Just a shutout for folks to know that we love them :D 14:33:56 Here it goes: 14:34:19 From a bug squashing POV, I'd like to give a special thanks to bunting and kairat 14:34:32 They have been working like machines on killing as many bugs as possible 14:34:39 and it's been amazing to review all those patches 14:34:41 bunting kairat ++ 14:34:46 Evil bugs :) Cheers! 14:34:52 thanks) 14:34:55 thanks folks 14:34:57 * nikhil claps 14:34:57 Please, do keep it up and sorry for all those -1's/-2's and nitpicks 14:35:15 Bare with us while we try to keep up with your work 14:35:30 I'd like to make one more recognition for this milestone 14:36:08 and it goes to mfedosin for his amazing work in making nova work with glance v2. It has not been easy and it's required a good deal of cross-project interaction, patience and tolerance 14:36:24 That's one of our priorities for Mitaka and it's moving forward quite well 14:36:42 So, thanks mfedosin again for your efforts there and please, don't give up. :D 14:36:52 oh... it's not done yet, actually 14:37:02 but will be very soon 14:37:07 I know it's not done, that's why I'm begging you to not give up 14:37:10 :D 14:37:12 and trusts too 14:37:32 I won't give up for sure :) 14:37:42 thank you :) 14:37:57 It goes without saying that the whole team did an awesome job for M-1 and the review rate has been way better! But, I think this 3 folks deserve a special thank from all of us this time :) 14:38:08 * flaper87 sends beers to everyone 14:38:13 moving on 14:38:15 actually 14:38:36 if you would like to recognize someone's work, please, feel free to add this topic to the agenda at any time 14:38:51 I'll do it once per milestone at least 14:39:04 ok, now, for realz, moving on 14:39:11 #topic CORS headers in glance: https://review.openstack.org/#/c/219478/ (kairat) 14:39:33 kairat: floor is yours 14:39:50 Ok, please look into review 14:40:08 krotscheck would like to clarify what headers are required for glance 14:40:43 but after analyzing the question it turned out that we have a lot of headers, because we are passing metadata as headers 14:41:11 So I am wondering what minimal set of headers we should set as default for CORS 14:41:11 (that's v1 only I think?) 14:41:11 kairat: that's v1 only, though. 14:41:13 AFAIR 14:41:15 :D 14:41:17 mclaren: yup 14:41:31 That's perfect 14:41:33 probably ok to support just v2? 14:41:39 +100 14:41:40 mclaren++ 14:41:45 mclaren, no objections here 14:41:49 yes 14:41:51 awesome, another happy customer. NEXT! 14:41:53 :D 14:41:58 cool 14:42:21 it looks fine 14:42:22 kairat: tsufiev I take you're all set now 14:42:30 still, just enabling CORS headers is not enough to solve the task of Glance images upload from Horizon (the one that CORS is going to solve) 14:42:36 flaper87, yep 14:42:43 flaper87, so, we're not ^^ :) 14:42:49 * flaper87 reads 14:43:07 tsufiev: what do you mean by "the task of glance images upload" ? 14:43:08 tsufiev, I guess the next topic is about that 14:43:16 "task" is a dangerous term in this team 14:43:22 flaper87, see https://blueprints.launchpad.net/horizon/+spec/horizon-glance-large-image-upload 14:43:25 :D 14:43:29 flaper87: topic change 14:43:31 the last comment in w 14:43:36 *in Whiteboard 14:44:14 I think we are talking about "multipart/form-data support in glance api " 14:44:22 nikhil, yes 14:44:23 not that I want to shun this discussion :) 14:44:29 flaper87: if we want to upload files to glance from Horizon we have to deal with multipart/form-data 14:44:30 nikhil: tsufiev ahhh 14:44:33 tsufiev: you hijacked the topic 14:44:36 bad bad bad tsufiev 14:44:40 :) 14:44:41 I don't know much about this tbh -- is there anything here to consider around image import? 14:44:45 #topic multipart/form-data support in glance api (tsufiev) 14:44:50 now we can talk about that 14:45:00 flaper87, sorry, I'm not familiar with local policies ) 14:45:09 tsufiev: are you aware of the efforts going into revamp the import process for images? 14:45:21 tsufiev: np, I was joking 14:45:53 I think this is something that will have to be considered as an enhancement for the new import process 14:45:55 flaper87, I had a discussion with Brian Rosmaita in spec CR comments, but according to him, Glance wants to offload this thing to Swift 14:46:11 and that's the point I disagree with 14:46:15 tsufiev: oh yeah, that sounds outdated 14:46:18 :P 14:46:29 discussions have evolved since then 14:46:36 and direct upload is not going away 14:46:51 flaper87, good to hear :) 14:47:01 Anyway, I think we'll have to take this under consideration for the revamped import process 14:47:04 mclaren: agreed ? 14:47:07 so, afair this is supported to an extent 14:47:10 Soo 14:47:16 perhaps not in mitaka but at least for N 14:47:20 hm... 14:47:23 oh wait, that's just reads 14:47:25 nvm 14:47:28 flaper87, is there any blocker? 14:47:30 Should we allow to upload image without task? 14:47:31 range writes is different 14:47:40 tsufiev: range writes is a complicated problem 14:48:01 because we need to deal multipart/form-data in this case 14:48:04 tsufiev: I had a discussion with swift PTL and he mentioned it to be a PhD thesis 14:48:09 tsufiev: yup, we're not going to modify the current upload path. That's our current blocker 14:48:15 flaper87: yup 14:48:46 we're almost at the end and there's one more topic 14:48:48 nikhil, flaper87: I heard that multpart/form-data decoder could be implemented as an additional middleware 14:48:56 I think it was ativelkov who told me that 14:49:11 tsufiev: we'll explore that for sure, it just needs to wait a bit longer 14:49:21 last write wins and when you enable multiple uploads, guaranteed success rate of a data write will be significantly low 14:49:32 as we don't have cycles to discuss/implement that and it isn't part of the mitaka priorities 14:49:46 ok, moving on! (sorry for cutting the discussion) 14:49:49 tsufiev: ok, we should discuss this offline. I am interested in knowing more 14:49:49 #topic SHA-2 and openssl dependency 14:49:54 no idea who added that 14:49:56 nikhil, ok 14:50:07 that's me 14:50:15 hemanthm: hey! :D 14:50:23 https://github.com/openstack/glance/blob/master/glance/common/signature_utils.py#L43-L48 14:50:27 * mclaren note to self: s3 objects seem to support multipart http://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectPOST.html 14:50:54 So, image signing makes use sha-2 exclusively for digest 14:51:02 bpoulos: ^ 14:51:23 And, SHA-2 is supported by default on openssl 1.0.1 or higher versions 14:51:51 now, on certain platforms like debian6, openssl only goes up to 0.9.8 14:52:02 ah mmh :( 14:52:02 which doesn't support SHA-2 by default 14:52:13 so, glance build is broken on those platforms 14:52:37 as dhellmann mentioned in the drivers meeting, can we pick the default based on the openssl version? 14:52:38 bpoulos: ^ 14:53:03 * flaper87 throws some cookies in the channel to grab bpoulos attention 14:53:06 does anyone recollect what metadata is set? 14:53:51 configurability might affect resiliency, so a bit curious 14:53:52 hemanthm: I'll take a look 14:53:56 I wasn't aware of this dependency 14:54:02 I knew cookies would work 14:54:08 :) 14:54:09 haha 14:54:13 bpoulos: thank you! 14:54:27 aaaaaaaaaaaaand another happy customer! 14:54:29 :D 14:54:41 can I get some cookies too? 14:54:48 #action bpoulos to look into the openssl dependency for the default signature method 14:54:57 * flaper87 gives hemanthm cookies and a cup of coffee 14:55:21 I'd have thrown coffee in the channel but that'd have been a disaster 14:55:25 #topic Open Discussion 14:55:29 o/ 14:55:57 flaper87, there is one more topic from bunting as I can see 14:56:00 Last week I asked for feedback w.r.t the glance ocmmunity, thinks that are working and especially, thins that are broken 14:56:15 Yeah 14:56:27 I was gonna bring up what we were talking about in the glance room earlier 14:56:30 I've got some feedback but not as much as I'd love too. Here's another request for feedback :D 14:56:42 oh, I'm sorry, I didn't see it 14:56:47 there are 4 mins left 14:56:57 is that enough? or should we wait for next week? 14:57:00 There might not be enought time 14:57:02 or discuss it in -glance ? 14:57:18 flaper87: may be carry over the discussion topics to next meeting (there are 3) 14:57:22 ? 14:57:35 yup, I'll do that 14:57:37 Can I get opinions on Bandit tests? Any particular tests that people think are important? 14:57:43 oh the Discussion Topics section 14:57:48 I totally skipped that 14:57:57 * flaper87 slaps himself 14:58:06 sorry about that 14:58:08 and I thought it was the time 14:58:18 :P 14:58:33 it was the time and my blindness 14:58:39 * flaper87 freaks out when he's running out of time 14:58:45 it's a dangerous situation to be in 14:58:48 Haha don't worry 14:59:00 It seems we had two topic sections 14:59:09 lets have a single section for all topics 14:59:16 we can use sub-bullets 14:59:38 ok, thanks folks 14:59:43 #endmeeting