14:00:03 #startmeeting glance 14:00:04 Meeting started Thu Sep 20 14:00:03 2018 UTC and is due to finish in 60 minutes. The chair is rosmaita. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:05 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:07 The meeting name has been set to 'glance' 14:00:10 o/ 14:00:12 erno is going to be a bit late 14:00:21 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:00:29 #topic roll call 14:00:55 we should skip the updates part for later 14:01:05 good idea 14:01:30 #topic periodic tips jobs 14:01:37 they're all green! 14:01:52 great 14:01:58 Woot 14:02:13 we still need to stop them from running in stable/rocky for glance and the client 14:02:32 andreas had some concerns about the change, but i did answer this morning 14:02:46 https://review.openstack.org/599837 14:02:46 just saw the patches 14:02:54 https://review.openstack.org/599844 14:03:23 yeah, please take a look and let me know what you think ... i'd really like to merge these or give them up to someone else 14:03:46 but enough about me 14:04:15 #topic remove bandit tests from stable branches 14:05:05 i guess the new thing there is backporting all the way to pike 14:06:20 yes 14:06:38 did anyone talk about this at the PTG? 14:07:09 this topic was not on the list for PTG discussion 14:07:12 I unfortunately had to miss all of the Glance sessions. :( 14:07:20 i'm sort of leery about turning stuff off in still-active stable branches 14:07:41 because i think imacdonn is going to make a bugfix backport to pike 14:07:46 At least with bandit, that's not likely to uncover anything in stable branches. Especially the older ones. 14:08:05 i'll admit i don't know much about bandit 14:08:10 If there are issues, we should catch them before they are backported. 14:08:21 Assuming someone actually looks at the output. 14:08:27 :) 14:08:33 :d 14:09:35 let's put this down for jokke to take a look at 14:10:00 #action jokke decide backport of https://review.openstack.org/#/c/598426/1 14:10:18 ok, looks like next topic is really for jokke, too 14:10:25 yes 14:10:30 #action jokke look at item 4 in today's agenda 14:10:54 We can just assign all actions to jokke since he's not here. :D 14:11:01 ok, next item is about something discussed at the PTG 14:11:17 #topic PTG report 14:11:32 abhishekk: could you give a quick rundown of any highlights from the PTG? 14:11:32 o/ 14:11:42 ah, jokke_ is here 14:11:48 abhishekk: you are saved by the bell 14:11:56 ++ 14:11:57 #chair jokke_ 14:11:57 Current chairs: jokke_ rosmaita 14:12:07 literally just got home ... on from the conservatory ;) 14:12:11 jokke_: we held off on the update for you 14:12:26 ok 14:12:28 also, you have some actions assigned to you 14:12:41 so the highlights are pretty much: 14:13:07 we need to focus our limited resources to stabilize things again for Stein cycle 14:13:40 Meaning we will be making the multiple back-ends stable api by S-3 14:14:11 the other thing we will be working on is the cluster aware glance 14:14:42 mainly making pre-caching images possible and image import with glance-direct without shared filesystem 14:14:58 other than that we will be pretty much doing bugfixing etc. 14:15:15 what do our resources look like for Stein? 14:15:34 we have a milestone targets written down at the end of etherpad https://etherpad.openstack.org/p/stein-ptg-glance-planning 14:15:35 Limited. :] 14:15:36 very thin 14:16:26 ok, thanks ... i have not been following the ML at all 14:16:42 so there might be other feature work coming given that there is actual people doing it 14:16:46 Luckily we have 2-3 new contributors willing to take care of some lite features 14:16:54 nice! 14:17:12 i just looked at that etherpad ... was it really just jokke_ and abhishekk at the PTG? 14:17:36 (sean said earlier that he had other obligations and missed the glance sessions) 14:18:00 on day one we were 3 people an day 4 we were total 4 peoples :D 14:18:08 yes, there was some interestand promises of people, but I'll be convinced when I see those people start to show up :D 14:18:21 s/day one/day 3 14:18:30 well, it's good that there were some new faces 14:18:35 yes 14:19:06 yes 14:21:32 (am I disconnected?) 14:21:41 no 14:21:42 abhishekk: i see you 14:21:47 ok 14:21:51 I just got browser up to get agenda open 14:22:04 so I guess that's it ... we can move on 14:22:24 #topic remove bandit tests from stable branches 14:22:51 we discussed this, decided that you need to decide if it should go into stable/pike 14:23:10 none of us have any objections 14:23:17 ok, will have a look, but should be fine dropping from there too 14:23:20 but it's up to the ptl 14:23:50 #topic backports 14:24:11 there was 2 backports listed ... something special about these? 14:25:09 ohh this is the location replacement thingie 14:25:49 jokke_: i think it's just the state transition change 14:25:53 at least at this point 14:26:17 i don't know why i am being such a weenie about these backports, it's another one i didn't want to approve myself 14:27:36 Yeah I'm not sure ... it definitely breaks our API backwards compatibility and I'm not still convinced it's a bug 14:27:50 but I don't know if it's worth of fighting 14:28:34 i do think it's a bug ... is the API break that now you'll get the state transition? 14:29:30 (just wondering if i am missing something) 14:29:32 well so far we haven't done state transition until you do add (which really is insert) 14:30:02 and I don't think it's documented anywhere either that the replace should trigger any state transitions 14:30:21 so in that perspective I think it's backwards breaking change 14:30:50 well, the lack of documentation is certainly true 14:31:01 i see your point 14:31:30 if it was documented as it should do it this would have been really straight forward bug to be fixed and backported 14:31:44 but that has never been the case 14:33:10 i don't want to leave imacdonn hanging on this, maybe something to the ML? 14:33:25 maybe operators list, just to see if it would break anything 14:33:42 my thought as well 14:33:46 makes sense 14:34:02 what smcginnis opinion 14:34:49 although there is really no reason why anyone should ever use the replace against queued image that has no locations 14:34:55 but I do see the frustration 14:35:10 Does someone have the review link? 14:35:24 ++ to your comment about no reason why anyone would do this 14:36:02 https://review.openstack.org/598204 14:36:08 smcginnis: ^^ 14:36:29 Oh right, thanks. 14:36:47 I took that as a bug fix, so even though it is a behavior change, it is a change to be "more correct". 14:36:58 smcginnis: my feeling exactly 14:37:10 though see what jokke_ said above about maybe it is not really a bug 14:38:21 so, it has merged into master 14:38:31 I can see it either way, so I will defer to the experts. :) 14:38:36 i wonder if someone should write up a release note about it being a bugfix? 14:38:37 But if it's merged to master... 14:38:45 then that way we can backport more easily 14:38:49 maybe without going to the ML 14:38:54 That may help. Probably should have had that with there being a change in behavior. 14:38:56 except to give a heads-up 14:39:16 (that last comment was about going to the ML) 14:39:16 so I'm not gonna push revert for the master patch but I'm not eager to backport it either 14:39:37 Maybe just leave it as is in stable and move forward? 14:39:43 A release note would still be good in master. 14:39:51 well, the one thing we know is that there is someone who needs to use this on the older branches 14:40:03 question is whether we should make them carry their own patch 14:40:24 smcginnis: that's what I was thinking ... but lets ask people's opinion for it ... I know it will make Ian's life easier if we backport it 14:41:01 ok, how about i work with ian on a release note, and then we send something to the ML? 14:41:29 yeah, lets not make the call hastily now. 14:41:48 #topic Validation data with location update, as discussed at PTG 14:41:54 we need the release note anyway, and this way there will be something definite to refer to 14:42:00 so this is related to the above 14:42:33 there is currently no way to set any of the checksums when the locations patch call is used 14:43:20 and we discussed options to allow that pre/during the ptg 14:43:29 did that change get merged into the API? i looked at a patch week before ptg, but then stuff came up and i never followed up 14:43:56 IMHO I think we should allow setting the checksums on the same call when the location is set and the image goes active 14:44:26 there is spec lite 14:44:29 #link https://review.openstack.org/597648 14:44:45 and allow it on no other calls, you mean? 14:44:47 api implementation 14:44:51 #link https://review.openstack.org/597368 14:45:05 and client implementation 14:45:08 #link https://review.openstack.org/602794 14:45:18 rosmaita: correct 14:45:45 looks like i did not leave comments, wanted to look more closely, but i liked the general approach in the patch 14:45:58 I don't think we should break the API that has been stable for the other use cases but I'm willing to make this happen as there is no other way to get the info in 14:46:16 well, it's only a teensy-tiny break 14:46:21 and last thing I want is the api node pulling the data down just to calculate checksum for it 14:46:34 right 14:46:52 and we don't want some clever person to start hiding checksum data in the location metadata 14:47:09 so i am glad that this patch will reject any such attempt, iirc 14:47:25 it's not teensy break for any other usecases as we would need to start rejecting the data and killing images again upon upload/import if we allowed it in all cases 14:47:43 i mean, they can still do it, but they can't call it 'checksum', 'os_hash_algo' or 'os_hash_value' 14:48:20 yeah, if you it into some other property it won't break anything if it's wrong 14:48:30 so you are still dead-set against adding checksum to v2 API image-create? 14:49:17 yeah, that would need to break way too much ... the fight was lost when the API was designed and it works as designed 14:49:35 (sorry I was busy in another meeting) 14:49:58 ok, i will not argue 14:50:18 specially now when we have the conversion in image import which means that you really can't trust that the user know correct checksum for the data that ends to the backend 14:51:14 that's a good point 14:51:23 and I don't want glance just overwriting the user given checksum either 'cause that would invalidate the whole point making it possible 14:51:51 so I'm ok allowing it in locations patch and limiting it to that 14:52:18 well, i do think using the location is an elegant way to smuggle it in for this particular use case 14:52:29 then it's really admin's responsibility to not break their cloud ;) 14:52:31 location metadata i mean 14:52:52 yeah, so lets continue the discussion on reviews 14:52:57 ok 14:52:59 so the current lite-spec is as per the discussion during the PTG 14:53:09 #topic Open Discussion 14:53:19 I have one 14:54:02 we need to add multiple backend support (support to read multiple backend config changes during scrubber and cache service start) 14:54:18 so should I mark it as a bug and propose patches against it? 14:54:48 abhishekk: if there is issues with those two, then yes lets open a bug and work it through that way 14:54:53 that makes sense to me 14:55:02 jokke_, rosmaita ack 14:55:19 it's clearly something we have overlooked and is part of stabilizing the feature 14:56:02 that's it, over to rosmaita 14:56:17 i have some news ... i haven't put up a stackalytics patch yet because i don't know my new email 14:56:30 but this is my last week at verizon 14:56:40 and actually, i am on vacation until friday 14:56:53 rosmaita: Good news, right? 14:56:55 i'm headed to Red Hat 14:56:59 Woot! 14:57:04 thanks! 14:57:07 \\o \o/ o// o/7 14:57:09 congratulations \o/ 14:57:18 thank you! 14:57:22 well that is good and not so good news 14:57:27 The gravitational pull is getting strong over there. :) 14:57:44 i have orientation early next week, so will mostly be offline until wednesday afternoon new york time 14:57:52 :d 14:57:59 yeah, the not so good part is that i'm being asked to work mostly on cinder 14:58:12 but i have been assured that i can participate in glance 14:58:12 Woot woot! :D 14:58:21 but it will most likely be mostly reviews 14:58:28 great 14:58:30 won't have time for glance coding 14:58:32 I'll help you with Cinder, you can help me with Glance. ;) 14:58:39 I will catch you for coding help :P 14:58:40 <3 14:58:40 smcginnis: deal! 14:59:19 we need to worry about diversity tag? 14:59:31 i think diversity tags were eliminated 14:59:35 last minute 14:59:36 so yeah, even less diversity and bit of shift of focus but I hope this is overall positive change (at least Cinder team will win hell of a lot) 14:59:37 great 14:59:37 so no worries! 15:00:10 ++, we can utilize rosmaita both ways :D 15:00:20 :D 15:00:34 ok, that's it for today. Thanks all, hopefully I'll be awake tomorrow to tackle those few things hanging 15:00:43 #endmeeting