14:00:03 <rosmaita> #startmeeting cinder 14:00:03 <rosmaita> #link https://etherpad.openstack.org/p/cinder-victoria-meetings 14:00:03 <rosmaita> #topic roll call 14:00:04 <openstack> Meeting started Wed Jul 1 14:00:03 2020 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 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:07 <openstack> The meeting name has been set to 'cinder' 14:00:19 <tosky> o/ 14:00:28 <walshh_> hi 14:00:35 <jungleboyj> o/ 14:00:55 <whoami-rajat> Hi 14:00:59 <e0ne> hi 14:01:31 <rosmaita> hello everyone 14:01:32 <rosmaita> #link https://etherpad.openstack.org/p/cinder-victoria-meetings 14:01:35 <geguileo> hi! o/ 14:02:04 <jungleboyj> Happy Wednesday! 14:02:12 <rosmaita> specs freeze in a few hours, so that's what we'll be discussing mostly today 14:02:18 <rosmaita> but first some announcements 14:02:25 <rajinir> o/ 14:02:33 <rosmaita> #topic updates - mid-cycle 14:02:44 <rosmaita> thanks to everyone who participated last week 14:02:53 <rosmaita> if you missed it, you can catch up on youtube 14:03:14 <rosmaita> though abishop said he was going to wait until after part 2 and then binge watch 14:03:20 <rosmaita> so, whatever works for you 14:03:21 <LiangFang> o/ 14:03:30 <rosmaita> #link https://wiki.openstack.org/wiki/CinderVictoriaMidCycleSummary 14:03:39 <lseki> o/ 14:03:40 <rosmaita> you can find a link to the recording ^^ 14:03:40 <jungleboyj> :-) 14:03:58 <rosmaita> #topic updates - monthly video meeting poll 14:03:59 <smcginnis> o/ 14:04:09 <rosmaita> #link https://rosmaita.wufoo.com/forms/monthly-video-meeting-proposal/ 14:04:57 <rosmaita> i noticed that there were some communication difficulties at the mid-cycle, so the poll includes an option about whether we should really do one meeting a month in video or not 14:05:20 <rosmaita> but if we commit to taking good notes in irc, i think it's worth a try 14:05:34 <rosmaita> anyway, the proposed first video meeting would be 29 July 14:05:41 <rosmaita> so we have a bit of time to think about this 14:05:56 <rosmaita> anyway, the poll closes 1200 UTC on Thursday 9 July 14:06:19 <rosmaita> #updates - brocade FCZM driver 14:06:25 <rosmaita> oops 14:06:36 <rosmaita> #topic updates - brocade FCZM driver 14:07:00 <rosmaita> i sent out an email summarizing what we discussed at the mid-cycle 14:07:05 <rosmaita> #link http://lists.openstack.org/pipermail/openstack-discuss/2020-June/015692.html 14:07:12 <rosmaita> so, we'll see if anyone cares 14:07:30 <rosmaita> #topic updates - need to keep an eye on the cinder-plugin-ceph-tempest job 14:07:41 <rosmaita> this was really bugging me yesterday, but maybe it's a fluke 14:07:46 <tosky> rosmaita: master or previous branches? 14:07:51 <rosmaita> we made the job voting on 11 march this year 14:07:55 <rosmaita> tosky: both 14:07:56 <tosky> there is an open review that should fix the issues 14:08:09 <rosmaita> #link https://zuul.opendev.org/t/openstack/builds?job_name=cinder-plugin-ceph-tempest 14:08:16 <rosmaita> tosky: that's good news 14:09:07 <rosmaita> #topic updates - code coverage job 14:09:37 <rosmaita> i finally followed up on my PTG action to put up a job so you can have the code coverage report in groovy html on each patch 14:09:47 <rosmaita> #link https://review.opendev.org/#/c/738687/ 14:10:00 <rosmaita> (it still needs reviews) 14:10:13 <rosmaita> right now, all it does is post the coverage output for you to look at 14:10:56 <rosmaita> we could add something that would pass/fail based on coverage falling past a threshhold 14:11:03 <rosmaita> e0ne is in favor of that 14:11:11 <whoami-rajat> #link https://6d4b1c17bd5712b1da1e-24da3718548989a700aa54b9db0ff79c.ssl.cf2.rackcdn.com/738687/8/check/cinder-code-coverage/7bb5c35/cover/ 14:11:13 <whoami-rajat> looks great 14:11:37 <e0ne> I know that not everybody agrees with me, so I put +1 on that patch 14:12:15 <smcginnis> Putting a pass/fail on coverage is a hard thing to do right. 14:12:28 <smcginnis> I've been in endless debates about what the right amount of coverage should be. 14:12:29 <jungleboyj> This is cool. Good info at least. 14:12:38 <rosmaita> yeah, it's a start 14:12:56 <rosmaita> i'll think about how we could track the coverage to address e0ne's concern 14:13:17 <rosmaita> my worry is that i don't want it to be non-voting and turn into the pylint job 14:13:29 <rosmaita> that's always red and so you don't notice anymore 14:13:44 <e0ne> rosmaita: +1 14:14:13 <smcginnis> Alternatively, we could just add coverage as part of the base commands that are run so it's always available for the normal UT runs. 14:14:56 <e0ne> rosmaita, smcginnis: we can use --fault-unnder (https://coverage.readthedocs.io/en/coverage-5.1/cmd.html) option 14:15:37 <rosmaita> let's discuss on the patch, or in a few meetings 14:15:49 <rosmaita> i think for now just having the info available is good 14:16:06 <e0ne> Agree. it's good for the beginning 14:16:11 <rosmaita> that's it for updates, unless anyone else has an announcement 14:16:13 <jungleboyj> ++ Starting by having it out there is a good step. 14:16:42 <jungleboyj> Don't want the voting question to stop that. 14:16:59 <rosmaita> yeah, let's make it a separate discussion 14:17:14 <rosmaita> #topic minor driver refactoring 14:17:28 <e0ne> rosmaita: thanks for raising this up! 14:17:30 <rosmaita> #link https://review.opendev.org/#/c/656888/ 14:17:37 <rosmaita> e0ne: thanks for doing it! 14:17:46 <tosky> e0ne: --fault-under is not good IMHO, it should be a delta 14:17:48 <e0ne> a minor change with a huge diff 14:17:56 <rosmaita> ok, what's going on is that e0ne did some refactoring 14:18:08 <rosmaita> and, it turns out that he didn't have to modify a lot of unit tests 14:18:12 <rosmaita> because they weren't there 14:18:21 <rosmaita> so my point in bringing this up is ... 14:18:39 <rosmaita> if you are a driver maintainer, look at that patch ^^ and check your driver 14:18:56 <rosmaita> and you might want to think about a follow up patch adding a test or two 14:19:05 <rosmaita> there are some sample tests on that patch 14:19:34 <e0ne> just for the note: I'm not going to add unit-tests for drivers for this patch or as a follow up patch 14:19:40 <rosmaita> the main thing is, we will review the patch by hand, but it's easy to miss something 14:19:45 <rosmaita> e0ne: good note 14:19:54 <rosmaita> yeah, this is a driver maintainer's responsibility 14:20:08 <rosmaita> so please take a look 14:20:26 <smcginnis> ++ 14:20:31 <rosmaita> but do it soon, like within the next few hours 14:20:57 <rosmaita> because it looks good, and we will merge it soon so e0ne isn't in merge conflict hell 14:21:14 <e0ne> :) 14:21:40 <rosmaita> #topic victoria spec review 14:22:03 <rosmaita> ok, i got sidetracked on some stuff, so my knowledge of the state of specs is from last night 14:22:10 <rosmaita> but here we go 14:22:34 <rosmaita> #topic spec - Remove quota usage cache 14:22:42 <rosmaita> #link https://review.opendev.org/#/c/730701/ 14:22:53 <rosmaita> Sam left some performance info on the spec, i think 14:23:24 <rosmaita> looks like Lucio's objection is formatting 14:23:38 <rosmaita> (which is fine, we want to be able to read the thing) 14:24:40 <rosmaita> so, at a glance, it looks like getting the usage dynamically isn't too big a hit 14:24:52 <jungleboyj> Looks like the performance impact is sufficiently minimal though. 14:25:26 <rosmaita> i suspect there will be some weird things that we won't know about until someone actually tries to implement this 14:25:41 <rosmaita> so probably no point trying to work them out in advance 14:25:42 <eharney> IMO we should look at our db schema and see if we have indexes etc. there that would make this perform as well as possible 14:26:01 <rosmaita> eharney: that is a good point, please add it as a comment to the spec 14:26:30 <rosmaita> yeah, good indices are key 14:27:19 <rosmaita> so, the author is in australia, and i think today is almost over there 14:27:55 <rosmaita> i think it would be ok to stretch the deadline for him to attend to some changes 14:28:14 <rosmaita> i think the mood here is mostly positive toward this spec? 14:28:22 <e0ne> rosmaita: +1 14:28:49 <jungleboyj> +1 14:29:33 <rosmaita> i guess this spec-freeze is a bit early relative to the ptg, but we can at least weed out the no-chance specs 14:29:56 <rosmaita> ok, i'll put a note that he can revise and it will still be acceptable until the next cinder meeting 14:30:13 <rosmaita> #topic specs - Support modern compression algorithms in cinder backup 14:30:23 <rosmaita> #link https://review.opendev.org/#/c/726307/ 14:30:45 <rosmaita> this one is kind of in the same boat, revision wise 14:31:04 <rosmaita> my feeling is that we should update our compression algo options 14:31:28 <rosmaita> but to what ... i think i agree with eric that we should just pick 1 of the 2 options 14:31:36 <rosmaita> as to which one ... 14:31:40 <eharney> yeah 14:31:50 <rosmaita> we're going to have to get this past the requirements team 14:31:58 <eharney> i haven't done much looking into which would be a better choice 14:32:16 <rosmaita> so i think in preparing the info for the patch for global-requirements, we can decide which is best 14:32:22 <eharney> i assume any would pass the requirements concerns 14:32:26 <e0ne> what about make these things plugable? 14:32:59 <eharney> they are 14:33:11 <eharney> but we don't want to give people a bunch of extra options just because we can, and have to maintain that forever 14:33:39 <rosmaita> i like zstandard better, but i'm not sure it's packaged as widely as the other option 14:34:06 <rosmaita> i think packaging is an issue for inclusion in global-requirements 14:34:34 <rosmaita> anyway, the author already has a patch up adding the algos to cinder 14:34:42 <rosmaita> so either one will be an easy code change 14:34:53 <rosmaita> it's really a matter of working out which is best 14:35:23 <rosmaita> so i think give him another week to get an argument together, and onto the spec? 14:35:57 <rosmaita> any objection? 14:36:08 <e0ne> eharney: actually, all algorithms and compressors are hard-coded 14:36:35 <e0ne> we can make it extendable like we did for db provider 14:36:42 <eharney> why? 14:37:03 <e0ne> so end-users will be able to add out-of-tree compressing algos to use it with cinder 14:37:21 <eharney> and then end up with backups that can only be consumed by certain versions of cinder on certain clouds on certain platforms.. 14:37:50 <smcginnis> ^ 14:37:53 <eharney> i don't see a benefit to justify that mess 14:37:55 <e0ne> goor point 14:38:08 <rosmaita> yeah, let's keep these as 2 separate issues 14:38:14 <jungleboyj> ++ 14:38:23 <rosmaita> so for now, we can say let's add a modern compression algo 14:38:38 <rosmaita> and we can discuss later whether the plugin approach makes sense 14:38:44 <geguileo> would it make sense to update the default compressor? 14:39:03 <geguileo> if we are going to add it to the requirements already and it won't be optional 14:39:33 <rosmaita> no, i think gzip is a venerable compression algo and we can honor it by keeping it the default 14:39:43 <geguileo> lol 14:40:11 <rosmaita> also, it's a standard library algo 14:40:12 <e0ne> geguileo: we should not break current deployments with rolling upgrades 14:40:49 <jungleboyj> It seems like this is a new feature that people can use if they wish. 14:40:56 <rosmaita> #topic specs - Reset state robustification 14:41:00 <geguileo> e0ne: that's a very good reason 14:41:08 <rosmaita> #link https://review.opendev.org/#/c/682456/ 14:41:14 <jungleboyj> Mmmm robustification 14:41:25 <e0ne> geguileo: that's why we still have tgt as a default option :( 14:41:37 <rosmaita> i think this is good, the open question is how to handle the --force option 14:41:40 <geguileo> e0ne: ouch 14:41:57 <eharney> i prefer just making the API safe and leaving the --force cases to cinder-manage 14:42:31 <rosmaita> i am ok with that 14:42:59 <rosmaita> eharney: why don't you revise the spec and people can comment, you can have a 1 week extension too 14:43:16 <eharney> ok 14:43:18 <rosmaita> #topic specs - Default volume type overrides 14:43:27 <rosmaita> #link https://review.opendev.org/#/c/733555/ 14:43:48 <rosmaita> this one had 2 +2s, i just left it open so people could still comment 14:43:58 <whoami-rajat> it looks perfect now 14:44:25 <whoami-rajat> thanks geguileo for the quick update 14:44:41 <geguileo> np 14:44:54 <rosmaita> ok, i will re-read and we can get this one approved before the deadline 14:44:55 <rosmaita> yay! 14:45:11 <rosmaita> topic specs - Backup Backends Configuration 14:45:20 <rosmaita> #link https://review.opendev.org/#/c/712301/ 14:45:26 <rosmaita> this one looks good too 14:45:49 <e0ne> thanks everybody for your feedback on this! 14:46:14 <rosmaita> looks like e0ne has revised, i think we can get this approved in the next hour or so 14:46:29 <rosmaita> #topic specs - Support revert any snapshot to the volume 14:46:30 <e0ne> :) 14:46:40 <rosmaita> #link https://review.opendev.org/#/c/736111/ 14:47:05 <rosmaita> smcginnis: thanks for your comment on there, i think you explained clearly why we want this to have a generic implementation 14:47:25 <rosmaita> that can be overriden by backends that can do better 14:47:31 <eharney> i think this spec needs more thought on the implications of doing this... it seems like a minefield of usability and perf problems 14:47:48 <smcginnis> I hate to be negative about it, because I do think it would be very useful. But there are many issues with trying to do this for all backends. 14:48:01 <eharney> it also doesn't explain how it actually works 14:48:09 <smcginnis> Very light on details. 14:48:27 <smcginnis> The work item list of "change restriction" is a little vague. :) 14:48:29 <rosmaita> ok, i think this one we ask them to continue working on for W 14:48:38 <rosmaita> smcginnis: i think W has a name? 14:48:48 <smcginnis> Wombat IIRC. 14:49:00 <rosmaita> glad i asked, i was thinking Wallaby 14:49:17 <rosmaita> ok, i will leave an encouraging comment on the spec 14:49:26 <smcginnis> Oh, maybe you're right. 14:49:31 <jungleboyj> I thought it was Wallaby. 14:49:34 <smcginnis> I really should know since I ran it, but now I forget. 14:49:36 <jungleboyj> I am pretty sure it is. 14:49:44 <smcginnis> Yeah, you're all right. Don't listen to me. :) 14:49:45 <eharney> this needs some details on what actually happens on a typical backend that can do this in the "ideal" way 14:49:46 <jungleboyj> Because that was the one I liked. 14:49:57 <smcginnis> eharney: ++ 14:49:58 <rosmaita> i was so devastated by my V name defeat, that i can't remember anything anymore 14:50:24 <eharney> because it's very unclear to the newer snapshots if you revert to an old snap 14:50:29 <eharney> unclear what happens to* 14:50:32 <jungleboyj> You weren't Victoriaous ? 14:50:44 <rosmaita> jungleboyj: :P 14:51:07 <smcginnis> eharney: I know at least for some it's an internal update of a pointer, but that's definitely not how all storage works. 14:51:13 <jungleboyj> This feels like one of those proposals that is a good idea but ends up being a dangerous minefield upon implementation. 14:51:19 <rosmaita> eharney: keep your thoughts coming, i will pull them out of the meeting log and put them on the spec 14:51:27 <rosmaita> ok, one final spec 14:51:40 <rosmaita> #topic specs - volume list query optimization 14:51:50 <rosmaita> #link https://review.opendev.org/#/c/726070/ 14:52:01 <rosmaita> i think you may have seen my note to the ML 14:52:10 <rosmaita> i think this is really a bug 14:52:18 <rosmaita> or actually a set of 3 or so bugs 14:52:51 <rosmaita> but, it may still be worth a spec 14:52:53 <eharney> i think rosmaita teased out some good details on this that i wasn't able to quite grasp when we last discussed this 14:53:01 <rosmaita> becuase the fix will have to be in 2 places 14:53:04 <eharney> but i haven't followed up enough on picking through this yetr 14:53:05 <eharney> yet* 14:53:10 <rosmaita> rest api layer and also the db layer 14:53:38 <rosmaita> but i don't think it's going to require a mv change 14:54:55 <eharney> i would suggest not calling this an "optimization" 14:55:08 <rosmaita> yes, the title needs to change 14:55:31 <rosmaita> it's basically "correct the volume-list filtering" 14:55:42 <rosmaita> and maybe "correct the volume list" 14:56:13 <rosmaita> and maybe "never, ever do this kind of change again where we hijack the volume status to have internal states that aren't displayable in the rest api" 14:56:23 <eharney> lol 14:57:22 <rosmaita> i think maybe i will write this up as a bug, and tell the author to feel free to pick up the bug? 14:58:22 <jungleboyj> Two Minute warning. 14:58:26 <eharney> sensible enough 14:58:39 <rosmaita> ok, works for me 14:58:53 <jungleboyj> ++ 14:58:57 <rosmaita> ok, cool, we got through all the non-WIP specs 14:59:05 <rosmaita> #topic open discussion for 60 seconds 14:59:10 <jungleboyj> *celebration* 14:59:25 <tosky> regarding the cinder-plugin-ceph-tempest test, this patch may help on train and older branches, but it needs to go through master and ussuri first: https://review.opendev.org/#/c/738273/ 14:59:29 * tosky was ready 14:59:37 <rosmaita> ty tosky 15:00:11 <rosmaita> already has a +2! 15:00:16 <rosmaita> ok, gotta go 15:00:19 <rosmaita> #endmeeting