14:01:39 #startmeeting Glance 14:01:40 Meeting started Thu Oct 15 14:01:39 2015 UTC and is due to finish in 60 minutes. The chair is flaper87. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:01:41 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:01:42 o/ 14:01:44 The meeting name has been set to 'glance' 14:01:44 o/ 14:01:47 o/ 14:01:49 0/ 14:01:51 hey folks! welcome :) 14:01:53 o/ 14:01:59 o/ 14:02:18 awesome, we have quorum. I believe jokke_ is around too! 14:02:22 let's get started 14:02:27 #topic Agenda 14:02:30 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:02:44 as usual, our agenda is there and we have some interesting topics to discuss 14:02:59 #topic Artifacts updates 14:03:04 ativelkov: nikhil mfedosin 14:03:08 news? 14:03:09 hi there! 14:03:20 hi 14:03:28 Alex wrote a big message 14:03:39 o/ 14:03:43 [openstack-dev] [app-catalog] [glance] [murano] Data Assets API for App Catalog 14:03:46 o/ 14:03:49 the email w.r.t the app catalog ? 14:03:50 ah yeah 14:03:57 planning to get to it tomorrow 14:04:25 I haven't read it yet, but it's about next implementation of Glance v3 API 14:04:44 ok, I'll read it carefully and I'd like to encourage everyone to do that 14:04:52 flaper87, we're waiting for your response on that matter 14:05:18 sure thing 14:05:22 flaper87, everyone is welcome in the discussion :) 14:05:32 thanks 14:05:49 I'd love to hear others opinions there. I think it's important 14:05:54 ok, moving on 14:06:00 #topic Drivers updates 14:06:10 We discussed a, hopefully, improvement for proposing small specs. These specs have been refered to as lite specs for a while and we didn't have a good way to manage them. In the hope of keeping the user's process simple, we've discussed using LP whislist bugs as lite specs and to implement a small process for it. 14:06:11 The discussion happened here: 14:06:14 #link http://eavesdrop.openstack.org/meetings/glance_drivers/2015/glance_drivers.2015-10-13-14.01.log.html 14:06:15 and I've started to document the process here: 14:06:17 #link https://review.openstack.org/#/c/234653/ 14:06:19 The document needs some work and jokke_ has stated he doesn't like the idea at all, very clearly. (at least he did it in lowercase :P) 14:06:21 I'd like to encourage everyone to go through the logs and the proposal. It'd be great to get some further feedback on the process. 14:06:23 Personally, I believe this is worth a try and it's definitely something we can revert or move away from if it doesn't work out. 14:07:18 I think it'd be healthy to keep this discussion on that review but, if there are questions already, we can take some time now 14:07:52 The meeting logs are not very long and the docs could use more eyeballs 14:07:55 :) 14:08:18 concerns? Shall we move on to the next update ? 14:08:19 :D 14:08:24 What will we do with the current bugs marked wishlist? 14:08:24 It doesn't hurt to try a beta 14:08:38 bunting: those will be triaged 14:08:42 meaning lets pick some potential lite-specs and see how people give feedback 14:08:54 yup, that's the goal 14:08:56 ++ 14:08:59 and then document the process 2-3 weeks after the summit 14:09:04 if it doesn't work, we can move to something else 14:09:23 o/ 14:09:25 or mention in the doc that we are trying 14:09:32 I will comment 14:09:41 nikhil: ++ 14:09:53 I did mention it's being introdued in Mitaka 14:09:57 introduced 14:10:03 I think it's gonna be really messy 14:10:22 but my comments are mainly on that doc change 14:10:50 ok 14:10:52 lets move on 14:10:56 #topic release updates 14:11:00 jokke_: floor is yours 14:11:19 So Glance got nice and shiny "11.0.0" Liberty tag just 35min ago 14:11:35 yay 14:11:36 w00000h00000 14:11:41 Thanks everyone and big hand to nikhil for being on the wheel 14:12:06 also stable/kilo point release was tagged earlier this week 14:12:15 w000h000000 14:12:35 speaking about stable: glanceclient stable/juno gate works now 14:12:37 thanks for the help keeping our stable together as well for all the bugfixes and backporters! 14:12:45 flaper87: gr8!!! 14:12:55 flaper87: yeah, there were some inconsistencies in stable/juno for client 14:13:17 and the project mark for py3 support was partial so, mostly incorrect 14:13:23 back to green now, which is good now that juno will soon reach EOL 14:13:49 * flaper87 hates EOL'ing versions that don't have a working gate 14:13:57 yeah ... I'll try to have a look if there is something critical we should backport for that before closure 14:14:05 anyone are you aware of the deprecation dates for juno? 14:14:22 it's likely going to be decided at the summit 14:14:23 yeah, I don't remember the official deprecation dates 14:14:28 I see 14:14:30 nikhil: Iirc in 3months time or so ... the support for Juno was promised 15months 14:14:48 yeah, but there have been changes (or at least there were in YVR) 14:14:50 gotcha, thanks jokke_. but that's still a few months down .. 14:15:10 yeah we've had hope to keep bit overlap 14:15:10 there's an email thread about this, I should look it up 14:15:29 that's it from release front 14:15:32 ok, moving on if there's nothing else on this topic 14:15:34 good 14:15:42 #topic cross-project liaisons 14:15:49 #link https://wiki.openstack.org/wiki/CrossProjectLiaisons 14:16:00 There are a couple of things here: 14:16:26 1) We need to update that wiki page, namely, we need volunteers for the several liaison roles. 14:16:42 I'll be pinging folks that are listed there to make sure they still have time for it 14:16:42 I can be Glance<->Documentation liason, btw :) 14:17:03 BUT!if you want to volunteer for something, please, let me know. Even if there's someone listed there 14:17:06 mfedosin: awesome 14:17:07 since I work with Olena 14:17:19 #info mfedosin wants to volunteer for the glance->docs liaison role 14:17:23 I'm happy to keep the Stable 14:17:24 i think i already am cross-doc liason 14:17:26 I'll volunteer for security 14:17:43 w000h0000 14:17:47 many volunteers 14:17:54 keep it coming 14:17:57 we need one more 14:18:01 security for me too if there's room 14:18:21 rosmaita, ah, okay. you are already there 14:18:29 nova<->glance needs update and I think flaper87 is the the POC ? 14:18:35 we need someone that can help attending the cross-project meetings that may/may not disappear in the future 14:18:37 mfedosin: let's talk at the summit 14:18:45 nikhil: yeah 14:18:49 flaper87: I was gunna suggest that 14:19:15 I would attend those but they happen in a really bad time for me and I'd love to have some help there 14:19:18 planning to help with CPL meetings in Mitaka. 14:19:27 I know nikhil has attended some and <3 for that 14:19:30 flaper87: I can also share the workload on the release front if needed. Have been doing it (maybe too much) already ;) 14:19:35 rosmaita, sure! frankly speaking I don't want to get another responsibility :) 14:19:46 jokke_: sweet, would love to work with you on that 14:20:20 nikhil: mind if to take the CPL meetings one instead of security (or with security) 14:20:21 :D 14:20:24 I can do infra+qa if there's no spot on secuity 14:20:44 flaper87: sure, I figured. the second one should be PTL 14:21:09 trying to find a big difference between infra and qa atm 14:21:50 flaper87: so, what's the difference between CPL and inter project liaisons? 14:22:04 nikhil: you mean the CPL should be just for PTLs? I guess that's how it was presented but it's really hard for me to attend. I've participated in some but can't do them all 14:22:08 flaper87: is that just CPL meetings liaisons 14:22:09 ? 14:22:16 nikhil: yeah 14:22:22 Anyone knows who Nate Ziemann is? :P 14:22:28 I do 14:22:30 work with me 14:22:31 we'll clarify a lot during the summit 14:22:41 since there'll be a session on this 14:22:53 Seems to be our Product Working Group liaison, but never heard of him ... just wondering if he has put his name on wrong line 14:22:56 Ok, thanks everyone for volunteering, I'll add you to the wiki if you're not there already 14:22:59 flaper87: I don't mind attending CPL meetings as I do that for awareness anyways.. 14:23:11 we need to move on, many other topics to go through 14:23:28 nikhil: ++ lets talk offline 14:23:30 thanks, btw! 14:23:34 jokke_: he's prod mgr at IBM and active at the group there. I've had a chat with him in Liberty. He put his name them I sup.. 14:23:42 #topic Trusts in glance (mfedosin, kairat) https://review.openstack.org/#/c/233601/2 14:23:58 hey! 14:24:22 just wanted to inform you that we started trusts implementation in Glance 14:24:45 I would call it PoC at this stage) 14:25:07 here's the basic code https://review.openstack.org/#/c/233601/ 14:25:10 that's awesome 14:25:16 ++ 14:25:18 #link Trusts in glance (mfedosin, kairat) https://review.openstack.org/#/c/233601/2 14:25:21 ops 14:25:23 #undo 14:25:24 Removing item from minutes: 14:25:27 #link https://review.openstack.org/#/c/233601/ 14:25:43 and we're going to insert it in appropriate places in glance 14:26:03 do we have a spec for this? 14:26:07 mfedosin: do you have an ETA? 14:26:08 exactly in image uploading v1 and v2 14:26:10 nikhil: there's one 14:26:18 nikhil yes 14:26:19 I am trying to get a sense of what the scope is 14:26:19 flaper87, need to be updated 14:26:35 flaper87, I am going to update it tommorow 14:26:42 #link https://review.openstack.org/#/c/229878/ 14:26:45 got it, thanks! 14:26:46 ^^ the spec 14:26:49 :) 14:26:50 kairat, yes please :) 14:27:06 the problem was - how to pass trust into glance_store 14:27:13 please, keep the spec updated as we're in the process of digging into those 14:27:29 because there is no keystoneclient dependency there 14:27:36 mfedosin: looking forward to chatting about this at the summit 14:27:46 would it be worth it having a session on this? 14:27:49 there are free slots 14:27:54 there is session proposed 14:27:57 where? 14:28:00 did I miss it? 14:28:02 on the etherpad 14:28:13 but today after a discussion with Flavio we decided to add it as optional dependency to Swift backend 14:28:21 flaper87: late addition 14:28:21 deep blue just above x-proj 14:28:21 was that recently added? 14:28:22 flaper87, I added it 14:28:25 ahhh 14:28:31 ok, I'm not crazy then 14:28:38 I mean... 14:28:40 yesterday, I suppose 14:28:43 I am... but yeah 14:29:03 mfedosin: sounds great, I'll add it to the schedule 14:29:14 without trust we can't use multitenant swift driver 14:29:30 because it's unsecure 14:29:41 (see my latest cve) 14:29:43 awesome! 14:29:47 mfedosin: are you planning to add that as an optional way to authN? 14:29:51 so we should implement them asap I think 14:29:56 mclaren: it'd be great to get your review on the spec for this work 14:30:28 yup, I'm on it 14:30:34 mclaren: awesome, thank you 14:30:42 mfedosin: anything else? 14:30:48 because, with swift we still don't have a way upstream for operators to move away from bad old locations and prevent them from keeping swift auth v1. 14:30:52 nikhil, I'm not sure right now, but yeah - we can reathenticate with trust to get new token 14:31:15 ok, let's discuss this on spec. 14:31:18 thanks. 14:31:19 we can discuss it on the summit 14:31:24 ++ on spec 14:31:37 moving on 14:31:40 #topic Approach for location management policies (kairat)  https://review.openstack.org/#/c/233687/ 14:31:42 if you interested than put your name in the etherpad 14:32:22 So I created a spec that changes the policy approach for get_image_location 14:32:38 Now i see that light-spec should be different (will change it soon) 14:33:06 kairat: if you wrote the spec already, don't change it 14:33:07 The change proposes just exclude location info from server response when policy doesn't allow it 14:33:12 it's there, we can review that 14:33:12 flaper87, ok 14:33:30 I would like to have some feedback on it 14:33:47 kairat: hmm, are we not using show direct url config option? 14:34:08 Previously(in v2), we prohibited image-show, image-list of policy doesn't allow us 14:34:27 nikhil, when it is true then policy should manage this 14:34:31 IMO 14:34:33 yes we have config option for that and we support the policy limitations for to whom it's exposed 14:34:54 mmh, if we're adding policies, I wonder if we still need the config_option 14:34:56 Additionally, i found a bug https://review.openstack.org/#/c/233687/ 14:34:58 I'll comment on the spec 14:34:59 show direct url is off by default I think 14:35:05 mclaren: was just battling around this quite a bit so might be able to give bit more light there 14:35:11 may be we need it for v1 flaper87 14:35:18 (If I'm thinking of the right thing) 14:35:28 I can try 14:35:32 nikhil: oh, good point. I'll have to check 14:35:40 by default show multiple locations is off 14:36:04 but when it is true then get_image_location is useless right now 14:36:06 if you switch it on and then restrict the get_locations policy then lots of things don't work properly 14:36:14 because image-show image-list is also forbidden 14:36:23 mclaren, +1 14:36:24 It's bad, but hopefully a corner case 14:36:29 mclaren: ++ 14:36:33 there is bug open for that 14:36:43 you have to switch on two incompatible non-default things 14:36:57 yes 14:37:17 hopefully policies would be able to handle v1 too 14:37:18 Also I noticed that we cannot add/delete location is show_multiple is off 14:37:21 then we can deprecate it 14:37:27 nikhil: that's my hope 14:37:34 and showing image locations is very brave IMHO :-) 14:37:45 ++ 14:37:46 deprecating the option would be ideal so we can just use policies 14:37:49 well people want to do that mclaren 14:37:55 like ironic 14:38:10 not everyone runs a public cloud 14:38:19 * nikhil nods 14:38:21 and I'm a lot against feature adds on v1 ... if it hasn't been needed so far, we probably won't need it at the corner of deprecation either 14:38:25 mclaren, do you mean showing direct urls? 14:38:41 showing the location string 14:38:52 mfedosin, and this also, I heard Ceph also need this 14:39:01 'locations' is terrible 14:39:02 *this=location_utl 14:39:08 yeah, cinder probably do stoo 14:39:13 it can be a confusion hell 14:39:14 too 14:39:23 ok, we need to move on. Sorry for limiting the time on the topic 14:39:29 lets follow-up on the spec 14:39:33 ok 14:39:39 it should be a complete spec 14:39:42 not a lite one 14:39:45 kairat: thanks for the heads up and for putting the spec together 14:39:48 nikhil: agreed 14:39:49 but thanks kairat for already proposing a review 14:39:58 #topic Skip next week's meeting (flaper87) 14:40:07 I'd like to propose skipping next week's meeting 14:40:08 #vote yes 14:40:12 :P 14:40:15 :P 14:40:18 ) 14:40:28 I don't mind if the plane has wifi :P 14:40:40 unless someone screams in 3 ... 2... 1... I'll agree this 14:40:44 #vote yes 14:40:45 an hour of travel time burned some way ;) 14:40:48 :P 14:40:56 #vote yes 14:41:03 just pay for the full wifi plan and you're good 14:41:18 #agreed Skip next week's meeting October 22nd 14:41:18 flaper87: _if_ it exists 14:41:37 I don't care, but skipping something is always a pleasure 14:41:38 jokke_: man, don't get me started on airplane wifis 14:41:40 :D 14:41:45 :D 14:41:59 moving on 14:42:03 #topic Summit agenda (flaper87) http://mitakadesignsummit.sched.org/overview/type/glance 14:42:10 well, that's the agenda! 14:42:13 19min and a lot in the agenda still 14:42:17 I'll add trusts to it 14:42:31 I created a schedule for the summit based on the topics that were added in the proposals etherpad: 14:42:32 #link https://etherpad.openstack.org/p/mitaka-glance-summit-topics 14:42:34 At the top of that etherpad, I added the agenda for the summit and, as you probably noticed already, we have 2 free slots. jokke_ mentioned that we could take some time to do a heads up on DefCore in one of our working sessions. I think it'd be useful for the team but I'd also like to encourage everyone to join the cross-project session on this topic: 14:42:36 #link http://mitakadesignsummit.sched.org/event/402c14d0efed29d1abd21b51f6785762 14:42:38 Anyhow, the link to our schedule is: 14:42:40 #link http://mitakadesignsummit.sched.org/type/glance#.Vh-J4s-1RZI 14:42:42 If you have another proposal OR if you have some conflicts in the current schedule, please, let me know and I'll do my best to make you happy. 14:42:50 we now have just 1 free slot (trusts took one) 14:42:50 Are we looking for topics to fill some of the free slots? 14:43:12 The free slot could be used to talk about defcore, interop and glance 14:43:18 I'm not sure, but we could potentially talk about image formats? 14:43:32 if there are other proposals we could use that time for it 14:43:33 flaper87: are you planning to add tags for other projects to some of the topics? 14:43:41 or can discuss artifacts twice 14:43:45 nikhil: yes 14:43:51 cool 14:43:51 flaper87: mind to update the etherpad what you have put on the working sessions 14:44:10 or was it only the first one that had no details 14:44:11 the schedule is still a draft since I wanted to know if there were conflicts 14:44:22 jokke_: probably just the first one 14:44:23 it was ... sorry, my bad 14:44:26 the etherpad should be updated 14:44:28 :D 14:44:36 or, almost updated 14:44:57 again, session leaders, check your calendars and make sure there are no conflicts 14:45:06 I'll send an email out to the ML 14:45:16 but I thought about sharing it here as well 14:45:22 thanks to all the session leaders 14:45:26 we <3 y'all 14:45:34 at least I do :) 14:45:35 flaper87: the two late sessions on thursday conflict with searchlight 14:46:08 so maybe schedule stuff for people who don't care about searchlight at those times? 14:46:28 rosmaita: yeah, that's unfortunate. Everything conflicts this summit 14:46:43 i know, that's the way it goes 14:47:12 priorities ... everyone should know how to prioritize glance ;P 14:47:19 :P 14:47:22 rosmaita: I'll see what can be done but I kinda think those two should be there because the image import should go after the fishbowl session 14:47:34 don't worry 14:47:37 and we can't finilize the priorities before we talk about these sessions and other topics 14:47:45 I'll take another look there 14:47:58 i only want to go to the very last searchlight session at 5:20 14:48:01 think like a traveling salesman 14:48:21 so i will be at the image import follow up 14:48:29 mfedosin: next topic is yours but I'd like to give time to rosmaita's topic if you don't mind 14:48:42 we can talk about yours if there's more time at the end 14:48:52 oh, okay then 14:48:57 mfedosin: thanks 14:48:57 mine is real quick 14:49:00 #topic image import spec: https://review.openstack.org/#/c/232371/ (rosmaita) - 14:49:03 rosmaita: shoot 14:49:13 i'm thinking of doing a quick revision to handle the smaller points that have come up and put in some "note" for the topics that still require discussion, but a new patch set hides the inline comments ... any advice? 14:49:34 probably aimed mostly at people who have added comments already 14:49:53 there are also a lot in the regular comments section, so those won't be hidden 14:49:53 mmh, if the new PS is a must then do it, we can always go bac in history 14:49:55 works for me, I still have them pending in draft :P 14:50:00 not in time, which is unfortunate 14:50:04 rosmaita: go for it ... the comments will show up on the previous PS anyways ... make a note that people should read them before commenting the new version (as it won't be last anyways) 14:50:28 what normally happens is that people keep discussing in older PS until the discussion is over 14:50:38 I don't mind porting the comments as the new PS seems vital 14:50:42 which it never is but I digress 14:50:47 ok, sounds like new PS is OK then 14:50:53 rosmaita: I think we'll probably need a couple more revisions anyway 14:50:55 rosmaita: go go go go! 14:51:07 i'll work onthat righrt away and then we can send out to the ML for more comments 14:51:19 cool 14:51:25 rosmaita: awesome, looking forward to that 14:51:25 rosmaita: I think it's better to get the annoyances away so folks can focus on the bikeshedding :P 14:51:32 jokke_: ++ 14:51:44 jokke_: bike what? there's no such thing 14:51:55 jokke_: although, I'd rather call it something else 14:52:00 flaper87: I can introduce :P 14:52:15 I mean, bikeshedding is not descriptive enough 14:52:17 :P 14:52:31 ok 14:52:32 * jokke_ puts the Internet Troll hat on before reviewing 14:52:35 #topic Schema validation for 'image-list' in glanceclient (mfedosin) 14:52:51 mfedosin: 5 mins so we can have open discussion (sorry for the timebox) 14:52:53 I've got one question 14:53:10 some time ago we removed schema validation from image-list 14:53:15 https://review.openstack.org/#/c/211086/ 14:53:25 #link https://review.openstack.org/#/c/211086/ 14:53:38 do we need it back as an optional feature? 14:53:49 can we just work towards removing fully validating the server responses? 14:53:59 jokke_, ++ 14:54:01 jokke_: ++ 14:54:04 it should be either there or not, I don't see the value for it to be optional 14:54:06 but it changes our client 14:54:15 that was my recommendation as well 14:54:33 we shouldn't have it there. If there's a good use-case for it I'm all ears 14:54:43 we can do gating tests that validates the server responses against schema, but we don't need to gate it on every client run 14:54:51 so, scrolling back some history pages 14:54:59 the only thing I could think of was admin use, but I'm not sure how strong a case that is 14:55:01 but what about just 'get' a single image? 14:55:06 do we need it here? 14:55:16 it was introduced so that people can know that the image records they are getting back are consistent with what the clouds says they should be 14:55:37 mfedosin: I'd say no 14:55:41 and for cases when changes happen to images, they can be picked rather quickly 14:56:22 if the image-meta remains in such a fluid state and constrained by policy, then the user can't know what to expect at a given point of time 14:56:26 so, the final decision - we don't introduce optional feature for that :) 14:56:42 as the time bound expectation is rather dagenrous 14:56:43 am I right? 14:57:09 mfedosin: I'm happy to review any such proposals :P 14:57:10 SO, I think it 14:57:14 it is important 14:57:26 nikhil: as an option or always on? 14:57:35 always on 14:57:38 mclaren: but 14:57:41 nikhil, but we have cases when validation can fail 14:57:50 for example you changes list of supported images 14:57:59 well we have had so far nothing but problems from it 14:58:01 and one of formats were deleted 14:58:04 nikhil: not practical -- v1 can create images which don't comply with v2 validation 14:58:23 I'd love to hear examples 14:58:27 mclaren: ^ 14:58:34 check the client bugs :-) 14:58:40 as said I'd be pro gating schema validations but it does not belong to our client 14:58:44 there have been cases, fwiw 14:58:49 https://bugs.launchpad.net/python-glanceclient/+bug/1477910 14:58:49 Launchpad bug 1477910 in python-glanceclient kilo "A user can prevent another user listing images via v2 (in certain cases)" [High,Fix committed] - Assigned to Stuart McLaren (stuart-mclaren) 14:58:54 also we can't location when show_multiple_locations is false because of this 14:59:04 *can't add/delete 14:59:21 It just fails on client but server works correctly 14:59:24 there have been about 5 or 6 bugs I'm aware of with validation the cause 14:59:25 that said, I think we need to think this through. For example, I don't think v2 should return images that are not v2 compliant 14:59:39 nikhil: good example was the schema change that warranted RC2 (I think) 14:59:41 even if the image was created using v1 14:59:43 flaper87: agreed 14:59:56 ? 14:59:56 just a bug that blew our client without real reason 15:00:02 and we would have never caught it if not for validation jokke_ :) 15:00:06 it's a catch 22 15:00:22 ok folks, we're almost out of time 15:00:23 i agree with jokke_ that the place to verify compliance is the gate 15:00:25 mclaren: flaper87: how about opening a new official tag that lists such cases and we can discuss on those bugs there? 15:00:28 nikhil: that's why I'm proposing gating those schemas, not blowing our client on them 15:00:31 from a users point of view if the image is usable via v1 then v2 is broken if they can't use the same image 15:00:52 I'd like this discussion to move forward a bit more before creating a new tag 15:00:57 nikhil: there was a mail on the mailing list from Jay Pipes recommending it gets switched off (fwiw) 15:00:58 may be it should be a post-list process on client 15:01:01 There's a bug for client validation, lets talk there 15:01:09 or just create a new thread on the mailing list 15:01:22 flaper87, I'll write a message there 15:01:28 in mailing list 15:01:31 or discuss at next week's meeting :) 15:01:37 mclaren: I will look it up . thanks for the pointer. 15:01:40 mfedosin: thanks 15:01:42 LOL 15:01:45 ok, time's up! 15:01:45 rosmaita, which is skipped 15:01:47 thanks everyone 15:01:53 #endmeeting