14:00:19 #startmeeting Glance 14:00:20 Meeting started Thu Apr 30 14:00:19 2015 UTC and is due to finish in 60 minutes. The chair is nikhil_k_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:21 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:22 :-) 14:00:24 o/ 14:00:24 The meeting name has been set to 'glance' 14:00:27 Welcome all! 14:00:28 o/ 14:00:31 o/ 14:01:00 o/ 14:01:08 o/ 14:01:14 o/ 14:01:20 Looks like we have a decent turnout. 14:01:24 Let's get started 14:01:28 #topic Kilo tagged. 14:01:33 o/ 14:01:55 Kilo was tagged a few mins back. You may see an email soon. 14:02:05 o/ 14:02:16 w00000t, tags tags tags 14:02:19 o/ 14:02:25 o/ 14:02:28 o/ 14:02:37 \\o \o/ o// o/7 for kilotags 14:02:40 Any bugs/fixes need to be backported to stable/kilo and bring it to stable crew's attn ie Erno, Flavio, Zhi Yan and Nikhil (for Glance) 14:03:00 nikhil_k_: this is the final release tag, right? 14:03:00 The rules are more stringent so please keep that in mind. 14:03:05 yep 14:03:07 o/ 14:03:39 nikhil_k_: And that's the stable team for glance_store too, correct? 14:03:51 Correct sigmavirus24 14:03:55 and for py-client 14:04:17 right 14:04:21 :thumbsup: 14:04:34 There may be members from global stable team but these are POCs for Glance stuff. 14:04:54 * flaper87 wishes GH emojis would render in weechat 14:05:00 :-) 14:05:02 flaper87: make it so 14:05:10 flaper87: you could write a rust extension to do that 14:05:11 irssi? 14:05:20 lol 14:05:30 nikhil_k_: doesn't render them either :P 14:05:33 Anyways, questions? 14:05:33 nikhil_k_: you'd have to write that in perl ;) 14:05:43 * kragniz is glad it doesn't 14:05:46 :P 14:05:54 none from me 14:06:04 me neither 14:06:23 perl + UX ?= not like, dislike 14:06:40 Thanks 14:06:46 #topic Pre Summit discussions. 14:06:58 #link https://www.mail-archive.com/openstack-dev@lists.openstack.org/msg51762.html 14:07:47 Please read that if you did miss the note. We have a (very) early session planned on May 7th. That will likely overlap with meeting slot so no meeting next week. 14:08:05 I will send out a note on ML once things are confirmed timewise. 14:09:06 #action (everyone): Please suggest ideas that absolutely need to be discussed on 7th. Add them here: #link https://etherpad.openstack.org/p/liberty-glance-virtual-mini-summit 14:09:47 If you see 14:09:52 #link https://etherpad.openstack.org/p/liberty-glance-summit-topics 14:10:32 The topic "The direction of the Glance project" would most definitely be adjusted in the main event's schedule. So, I think we can discuss it later (on 12th or 13th). 14:11:05 For 7th, I would suggest topics that we think that may or may not be appropriate for main event. 14:11:20 But ideas are welcome. 14:11:55 Can we get some vote on the time schedule? 14:12:43 #startvote pre-summit sync on 7th from 1400UTC-180UTC 14:12:44 Unable to parse vote topic and options. 14:12:52 #startvote pre-summit sync on 7th from 1400UTC to 180UTC 14:12:53 Unable to parse vote topic and options. 14:13:13 Ok, let's vote without the bot 14:13:39 Who is against it? 14:13:51 nikhil_k_: are we sure we need three days? (Sorry if that's already been discussed/decided) 14:13:51 1800, not 180 there 14:14:08 Topics for May 7th would be the ones *not* discussed at the summit? 14:14:09 mfedosin: yes, thanks 14:14:10 1800UTC and bot will be fine :) 14:14:18 nikhil_k_: I'm not against it 14:14:22 mclaren: we do not. I was hoping that we wrap it up on 7th and 12th 14:14:32 ah, ok, thanks 14:14:38 13th is more of an alternative to 12th 14:15:03 ah, gotcha, thanks 14:15:10 Since summit is close around that time, we need to be alittle mindful of people's other obligations. That was mostly it. 14:15:30 thanks sigmavirus24 14:15:53 Ok, looks like plan for 7th sounds positive. 14:16:03 sounds good to me 14:16:12 the 7th might be doable, The last week before summit might not be so 14:16:15 nikhil_k_: do we want to reach out to the mongodb folks? See if they're interested in a session? 14:16:17 FYI, here's a link for timezone conversion http://everytimezone.com/#2015-5-7,120,cn3 14:16:19 Please do add your name in the list if not already. 14:16:31 nikhil_k_: Topics for May 7th would be the ones *not* discussed at the summit? 14:16:35 Thanks TravT , will add to etherpad. 14:16:55 marekd: I believe so 14:17:01 marekd: they would be the ones that are trying to make it to the main event 14:17:25 depending on the outcome of the conversations we would decide if to add to schedule or not 14:17:54 sigmavirus24: nikhil_k_: ok, and 12th? just a continuation, or some pre-summit discussions, as there will be no spot for them during the summit, right? 14:17:58 Sometimes, we have split sessions where two different topics share a session 14:18:07 nikhil_k_: yep 14:18:25 right 14:18:30 12th -> just a continuation, or some pre-summit discussions, 14:18:50 but they will/may have sessions that would also be discussed in the summit 14:18:57 nikhil_k_: gotit 14:20:17 Moving on 14:20:22 any option for may 6th, then? I might have a topic (added it to the etherpad) but it's unlikely i will be available next thursday 14:20:39 i will be on May 12, on the other hand. 14:20:57 #topic Open Discussion 14:21:15 marekd: I would encourage to discuss that now if possible (or something related to it) 14:21:58 This is the open discussion slot that you can freely use. If it looks positive then I will try my best to accomodate it. Rescheduling to 6th is expensive operation and I would like to avoid it. 14:22:12 nikhil_k_: ok, so let's not reschedule. 14:23:01 mfedosin: Re Trusts: I talked with Duncan Thomas from cinder. He looked at it and concluded it was doable(for cinder backup to swift), but never got around to actually doing anything 14:23:36 I think I can do it in glance 14:23:53 I'm going to make a spec and discuss it with you guys 14:24:10 There was a small bit of push back from keystone as trusts were not necesarily designed exactly for this use, but keystone admitted there was no real alternative IIUC 14:24:22 mfedosin: cool, loop me in 14:24:39 when you say you think you can do it, 'it' is trusts or some other mechanism? 14:24:43 yes, sure 14:24:55 trusts, of course 14:25:42 ok, I'll try to get Duncan to look at the spec, although he's hard to shepherd sometimes :-) 14:26:37 mclaren, okay, thank you and Duncan :) 14:26:45 np 14:27:54 Hi, so usually i hang out with the keystone folks, where i was focusing on identity federation in OpenStack. Since this is in a good shape i think there is a good moment to expand to other sevices. 14:28:30 One of the first non-keystone service could be glance, and an idea for 'intercloud image sharing' 14:28:43 I added the topic here: https://etherpad.openstack.org/p/liberty-glance-summit-topics (last one) few hours ago 14:28:51 so you likely haven't seen it yet. 14:29:27 marekd sounds interesting 14:30:08 I know folks are doing this already with some fairly hacky methods 14:30:35 mclaren: there is a side-project built on top of glance, glint. 14:30:55 There was a talk about one implementation @ Paris Summit, but haven't heard about that since 14:31:02 marekd: yes, I remember from one of the summits University of Somewhere (Vancouver?) 14:31:04 marekd: that was it I think 14:31:11 mclaren: Victoria 14:31:12 UBC 14:31:27 i thought CERN also showed something 14:31:39 TravT: same guys IIRC 14:31:45 TravT: not in this topic 14:31:53 https://www.openstack.org/vote-paris/presentation/glint-image-distribution-current-developments 14:32:12 yeah, that's it 14:33:36 still, glint relies on a concept of service accounts 14:33:54 i would like to followup and try to use federation 14:34:44 sounds much better, I was a bit frightened by what they were doing credentials wise 14:34:57 this would be somewhat identity part you guys probably don't care that much, but i feel there would be some glance work on that too. I added some use-cases at the bottom of my proposal in the etherpad 14:35:29 Sounds pretty interesting to me. 14:35:39 nikhil_k_: great to hear that. 14:35:51 mclaren: I think the glint use case seemed tied up to small single type of image. 14:36:10 Something that did not have prop data. 14:36:15 nikhil_k_: ok 14:36:46 just got Thierry's mail that Kilo is released \o/ 14:36:49 I think the architecture i am thinking about would be 'an appstore' like. So, one public cloud publishes some artifact (image, heat template) and others subscribe for some version of that (say always the newest) 14:36:57 (nice) 14:37:21 so the glint use case was that they get donated overflow resources from different cloud providers and they wanted to have up to date specific small image for their workload available on those clouds 14:38:05 right, and something that supported research on high perf data without affecting kernel performance 14:38:46 marekd: That sounds good. I think there may be _some_ overlap with what the community images followed by image marketplace feature of Glance 14:39:07 nikhil_k_: what do you mean? 14:39:09 Just want to make sure that we do not have two implmentations for a single use case 14:39:51 marekd: So the idea looks like a publisher + consumer model for certain type of images, right? 14:40:10 nikhil_k_: yes 14:40:11 Regulated by multi-tenancy? 14:40:13 kind of 14:40:49 The can be borrwed a little bit from the image-sharing concept that already exists today. 14:41:09 image-sharing in glance? From the user perspecive - yes. 14:41:10 We are trying to extend it to _really_ publish some images. 14:41:37 nikhil_k_: but still within one deployment of the cloud, right? 14:42:03 yes, right. Glance is region specific deployment (one DC == one universe) 14:42:20 at least typically 14:42:32 i could probably help making it publishable across trusted clouds 14:42:50 with some policy/limiting applies of course 14:43:13 The issue with federation of Glance is uuid collision. Although, that sounds rare, it actually happens in real world cases :-) 14:43:56 But yes, let's discuss this more. As not concerns were raised. 14:44:04 not many* 14:44:06 nikhil_k_: in federation you can freely draw a bold line in between the clouds 14:44:22 sure :-) 14:44:48 nikhil_k_: so, i wouldn't subscribe on uuids, rather...names, as crazy as it may sound. 14:44:51 or versions. 14:45:23 ok. 14:45:48 e.g i always want to be notified/have available image of the newest cernVM image 14:46:13 i trust cern, i am collaborating with them, and they now what and when needs to be published 14:46:21 they will take care of patching the image 14:46:34 i simply donated some spare cycles 14:46:57 gotcha 14:47:11 and the same may happen in commercial world 14:47:33 While we get ideas on the board, I want to quickly check if anyone else has somethign to discuss... 14:48:29 future of v1 14:48:40 jokke_: v1 must die! 14:48:45 ok, that looks important 14:49:01 marekd: Do you mind we catch up on this on May12/13 if possible? 14:49:03 can we finally make the call to deprecate it so there will be pressure to move the projects depending on it to v2? 14:49:15 nikhil_k_: of course! 14:49:19 marekd: thanks. 14:50:12 So, some folks pinged and sigmavirus24 yesterday on how difficult this v1, v2, tasks stuff is becoming. DefCore suggests v2. So, let's do it. Unless someone raised a red flag in the next 4 weeks. 14:50:19 me and* 14:50:37 s/rasied/raises/ 14:50:51 :thumbsup: from me 14:51:08 oh, I think kragniz was there too. 14:51:17 * jokke_ is more than happy to sacrifice some spare time to propose the change ;) 14:52:24 rosmaita: and I hope that death is extremely painful ;D 14:53:26 Anything else? 14:54:28 If not, thanks all! 14:54:36 thanks 14:54:45 Thanks 14:54:54 thanks, see you next week 14:54:57 #endmeeting