13:59:58 #startmeeting glance 13:59:59 Meeting started Thu Jul 5 13:59:58 2018 UTC and is due to finish in 60 minutes. The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:00 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:02 The meeting name has been set to 'glance' 14:00:04 #topic roll-call 14:00:06 o/ 14:00:06 o/ 14:00:09 o/ 14:00:13 o/ 14:00:14 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:00:14 o/ 14:00:49 I think we're good to go, everybody onboard :D 14:00:56 #topic updates 14:01:18 As expected nothing major has happened this week. 14:02:05 Couple of things to point out. Abhishek has the multi back-end patches up, lets get the spec amendment and the implementation reviewed and merged 14:03:06 Then he also got the migitation for OSSN-0075 up which will need reviews as well and the hidden images 14:03:19 About that tomoki has tested multiple backend with cinder and it is working, he has replied to my mail as well 14:03:32 ++ 14:03:54 I have the v1 endpoint removals close to done and it's currently passing all but functional tests but more about that later 14:04:12 i'll test out ossn-0075 after the meeting 14:04:35 #topic release updates 14:04:56 periodic test report 14:05:06 client, store are all green 14:05:18 the cursive test that was failing last week for glance is now passing 14:05:43 there's an intermittent timeout failure in glance-tox-functional-py35-keystone-tips 14:06:00 it has passed, failed, passed, failed, passed over last few days 14:06:20 #link http://zuul.openstack.org/builds.html?pipeline=periodic&project=openstack%2Fglance&job_name=glance-tox-functional-py35-keystone-tips 14:06:24 i will keep an eye on it 14:06:30 doesn't look like anything to worry about yet 14:06:46 upcoming release deadlines 14:06:50 k 14:06:58 glance_store Rocky release 18 July (1 day less than 2 weeks away) 14:07:05 so that is coming fast 14:07:08 did you check yet what is timing out there? 14:07:24 looks like a timeout during the tests 14:07:39 couldn't tell exactly which one 14:07:45 without further looking 14:08:06 ok, so the store release, like said I'd like to get the release out rather next week than the following 14:08:09 actually, it could be something else entirely 14:08:33 yeah, lets keep eye on that and if it keeps going dig in bit deeper 14:08:45 right 14:09:13 ok, and in other release news, the glanceclient final release is the week after glance_store deadline 14:09:24 and that is also the week for Rocky-3 milestone 14:09:53 that's about it 14:10:04 thanks Brian 14:10:10 #link https://launchpad.net/glance-store/+milestone/0.25.0 14:10:48 so Lets try to get the multi back-end store parts in, the we need to make the decision in which order we merge the glance stuff and what API versions we'll set 14:11:23 ok 14:11:27 ok 14:12:08 rosmaita, similar kind of blueprint need to be created in glanceclient? 14:12:20 possibly 14:12:35 ok 14:12:38 we're traditionally pretty cavalier with glanceclient 14:12:39 :) 14:12:48 :D 14:13:01 oh and please lets try to have the release notes ready around when the changes themselves are good to go so we don't miss any 14:13:28 i promised abhishekk that i will write up relnotes for this stuff 14:13:29 as usual we seem to be merging a lot at the last minute 14:13:39 rosmaita, +2 14:13:43 that's the glance way 14:13:45 ! 14:13:50 ;) 14:14:39 between we need to modify hide old images specs to reflect 400 to 409 if hidden column is not present 14:14:49 I will do that tomorrow 14:14:53 i'll do the relnote and admin doc for ossn-0075 while i;m testing later 14:15:16 sounds good to me 14:15:20 abhishekk thanks for remembering that one 14:15:29 Also I have added a check to restrict queued images to be updated with hidden (which is not present in the spec as well) 14:16:05 how do you mean? 14:16:25 abhishekk: good point 14:16:50 means when image is created and before uploading data to it if someone tries to update image with hidden = True then it should be rejected 14:17:26 makes sense 14:18:53 ok anything else? 14:19:11 no 14:19:33 #topic open discussion 14:19:41 I have couple of things here 14:20:21 1) Once I finally got rid of the obvious test failures I stumbled upon something and got abhishekk double check it for me today as well 14:20:40 so we have this cache management command for glance 14:21:20 which is still using v1 api so obviously no-one has reported using it as part of the v1 removal feedback what prevents them moving to v2 14:21:55 I think, for that reason, we should get rid of it as part of the v1 removal 14:22:10 it is broken for v2 as well, we need to get it functional 14:22:37 abhishekk: did you find actual v2 implementation for it? 14:23:25 Lack of feedback makes me wonder if anyone is actually using that. 14:23:26 jokke_, i went through code and found something related to v2 in auth.py 14:24:10 yeah, what I mean is that the command itself is v1 only implementation 14:25:41 so taken into account we have had v1 api disabled by default for quite a while and no-one has even mentioned it not working indicates to me that we should just remove it with the API and if there is use case for it we can do it for v2 right at that point 14:26:39 ok 14:26:54 smcginnis: my take on this as well. Honestly I did not even remember that the command exists 14:27:21 It sounds useful, but maybe it's not actually needed? 14:28:02 yeah, I'm not exactly sure how useful/meaningful manual cache management per api node is to anyone 14:28:26 kind of one of those things where you likely could use your time for something much more productive ;) 14:29:47 rosmaita: what do you think? 14:30:10 i don't have an opinion 14:30:41 i think it's a LRU ejection strategy 14:30:59 LRU? 14:31:12 kicks out least recently used 14:31:46 what did the cache manager do, exactly? 14:31:59 we have cache-cleaner, cache-prefetcher and cache-pruner as well 14:32:16 it just utility tool to show cached images or delete cached images 14:33:12 it allows you to list cached and queued images, delete images from cache and queue and add images to queue 14:34:08 so literally just micromanage your cache on that node 14:34:11 the cache-prefetcher was going to be useful for edge computing (or am i mis-remembering) 14:35:31 rosmaita: yeah, but it was never gonna be utilizing the cache_manage cli command (like said I did not even remember we had it) as it would be also touching only the local node glance service 14:36:08 well, there does not seem to be an outpouring of support for the cache manager 14:36:08 you don't want to be sshing in to every glance node you want precache images :P 14:36:19 that's what parallel ssh is for! 14:36:27 LOL 14:36:38 yeah, i didn't really mean that 14:37:06 so let me summarize what i think the situation is 14:38:05 the glance cache manager tool has a strong dependency on v1, which is being removed in rocky. we don't have bandwidth to mess with it and we also believe that it's not much used 14:38:24 so the current situation is that we will include a release note saying the tool no longer exists in rocky 14:39:10 that was my idea, but instead of just doing it I wanted to hear what you guys think first ;) 14:39:12 maybe the thing to do is to put up a spec-lite proposing removal, and send a link to the operator's list for comments 14:39:32 we could aim for removal in Rocky-3 milestone 14:40:09 I would just lump it in as part of the v1 API deprecation and removal and just drop it. 14:40:38 my concern is that it's not obviously a v1 thing 14:40:38 it's independent functionality 14:40:50 Although a release note would be good to have to make sure if someone gets surprised it is gone they can have a chance to find out what happened to it. 14:41:09 yeah, i agree that a release note is necessary 14:41:11 rosmaita: If you want note of it into the spec I can put amendment for it into the v1 removal spec on top of the reno 14:41:21 Yeah, it's independent, but just looking at an excuse not to have to deprecate it in rocky and remove in stein when that wouldn't make sense. 14:41:33 no-one reads either anyways but at least we could point that it's mentioned in multiple places :P 14:41:38 jokke_: That sounds like a good plan to me. 14:41:47 agreed 14:41:49 me too 14:42:25 gr8! 14:42:30 then the next thing 14:43:28 this has been hanging for long time, but I finally got abhishekk's test fixes encorporated to the patch that prevents us to launch the taskflow that we know will fail 557712 14:44:03 pretty well contained patch so would greatly appreciate your review efforts on it as well. abhishekk already went through it 14:44:16 yes 14:44:50 Do you have a link handy? 14:44:58 https://review.openstack.org/#/c/557712/ 14:46:31 that looks like a good change 14:47:31 looks like it's up to me and sean since you and abhishekk are co-authors 14:47:49 yes, that's why I brought it up 14:48:20 yes 14:48:27 if we end the meeting early, i will review it right now 14:48:30 :) 14:48:33 you guys have lots of stuff to review :D 14:48:53 well that's all from me 14:48:57 anyone else? 14:49:19 nothing from me 14:50:00 I need to put another patch to deprecate stores and default_store options 14:50:14 ok, going first 14:50:35 Nothing from me. 14:50:36 going second 14:51:07 sold! Thanks all! 14:51:12 #endmeeting