Wednesday, 2021-06-23

*** dviroel|out is now known as dviroel11:37
rosmaita#startmeeting cinder14:00
opendevmeetMeeting started Wed Jun 23 14:00:09 2021 UTC and is due to finish in 60 minutes.  The chair is rosmaita. Information about MeetBot at http://wiki.debian.org/MeetBot.14:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.14:00
opendevmeetThe meeting name has been set to 'cinder'14:00
rosmaita#topic roll call14:00
whoami-rajatHi14:00
simondodsleyHello14:00
rosmaita#action rosmaita - find out why rbd-iscsi-client is not mirrored on github14:00
fabiooliveirahi14:00
sfernandhi14:00
tbarronhi14:01
eharneyhi14:01
jungleboyjHello.14:01
abishopo/14:01
enriquetasohi14:01
rosmaitalooks like a good turnout, let's get started14:02
rosmaita#link https://etherpad.openstack.org/p/cinder-xena-meetings14:02
rosmaita#topic announcements14:02
rosmaitasince next week's meeting is the final meeting of June, it will be held in both video and irc14:03
rosmaitaalso, if you missed it on the mailing list, lseki has stepped down as a cinder core14:03
rosmaita#link http://lists.openstack.org/pipermail/openstack-discuss/2021-June/023206.html14:03
rosmaitaand, spec freeze is on Friday14:04
rosmaita#link https://releases.openstack.org/xena/schedule.html#x-cinder-spec-freeze14:04
sfernand:(14:04
e0nehi14:04
rosmaitahello ivan14:04
hemnahey14:05
rosmaitai'm going to change the order of stuff on the agenda14:05
rosmaitawe will discuss14:05
rosmaitasqlite (should take 2 minutes)14:05
rosmaita10 minutes on simon's black linter proposal14:06
rosmaitaand then the rest on spec reviews14:06
rosmaita#topic Drop support for SQLite < 3.714:06
rosmaita#link https://review.opendev.org/c/openstack/cinder/+/78887114:06
rosmaitabasically, all the currently supported platforms for openstack (centos-8-stream and ubuntu-focal) support sqlite >= 3.714:07
rosmaitawe have a check in the code for the case where someone is using an older version14:07
rosmaitaproposal is to remove it14:08
e0nesqlite is used only got tests, so it's ok to use any version which doesn't make our tests more complicated14:08
eharneythis only affects unit tests, so, not much risk14:08
e0neeharney: +114:08
rosmaitabut i was thinking maybe someone working with like queens, say, might be using an older sqlite on their dev box14:08
hemnaTrain ?14:09
rosmaitai think queens is still in -em mode14:09
rosmaitai believe train's supported platforms have sqlite >= 3.714:09
rosmaitabut i am not 100% sure14:09
hemnaI think train was ubuntu 18.0x > ?14:10
rosmaitaok, let's do this ... if anyone can think of a reason *not* to do this, please leave a comment on the patch before Friday 1200 UTC14:11
rosmaitaotherwise, i am inclined to go with e0ne and eharney on this14:11
rosmaitathanks!14:11
hemna+114:11
rosmaitaok, it is almost 14:12 now14:11
e0ne#link https://governance.openstack.org/tc/reference/runtimes/train.html14:11
rosmaita#topic discuss allowing `black` to be used as an auto-linter14:12
rosmaitasimondodsley: that's you14:12
e0neit's good to have it but we'll lost history14:12
simondodsleyi like to use `black` as an auto-linter - it makes life super easy  especially for seeing future changes14:12
simondodsleye0ne: I agree you loose history - how much of an issue is that?14:13
e0negit blame is usesul, but I prefer to have black :)14:13
simondodsleyIt isn't 100% perfect, but it makes linting somuch quicker14:13
rosmaitai think for the main cinder code, losing history is definitely an issue14:13
simondodsleyI have found a couple of weird things that `tox -e pep8` fails with when using `black` but they are very trivial14:14
e0neif we decide to use it, let's adopt it's settings to keep current code style14:14
eharneyi thought black was opinionated and didn't really have settings to adjust14:14
hemnaI've used black in other projects, and it doesn't always produce the most readable code IMHO14:14
simondodsleyI'm not saying it should be mandatory - yes it is opionated14:15
hemnaeharney it is very opinionated14:15
e0neeharney: it has some settings and/or ignore list14:15
rosmaitaso we can see the differnces:14:15
rosmaita#link https://review.opendev.org/c/openstack/cinder/+/792462/14:15
hemnaIt can produce stuff like this: https://github.com/craigerl/aprsd/blob/master/aprsd/messaging.py#L561-L57314:15
hemnawhich I think sux0rs14:16
eharneysee, about the first thing it changed on line 57 there is annoying14:16
eharney(breaking our standard formatting of help text on a new line after the option name)14:16
hemnaI use black in my ham radio python project called aprsd ^^ link above, as a mechanism to ensure that non python people can lint format their code before commit.14:16
eharneyalso not sure why it wants to change a lot of single-quoted strings to double quotes, weird14:17
simondodsleyFor new code is it a godsend, but true it does break some of the current standards, but are they critical?14:17
rosmaitayes, it definitely has its quirks14:17
hemnaI think the function definition 'formatting' really sucks14:18
eharneyi don't really see a benefit when pep8+hacking gets us quite consistent on style already14:18
rosmaitai like some of the stuff it does, but hate other things14:18
jungleboyjhemna:  ++14:18
rosmaitasimondodsley: we really want a consistent style to make it easy to read any of the code files14:18
jungleboyjeharney: ++14:18
simondodsleybut using `pep8` requires lots of manual intervention and that is a waste of productive time14:18
eharneyi just ran "black" on cinder/volume/manager.py and it blew up and told me to report a bug14:19
simondodsleyI can `blacken` code in 5 seconds whereas manally editing for `pep8` compliance takes ages. 14:19
eharney(was hoping to see what it did with type annotations)14:19
simondodsley`black` is a superset of `pep8` 14:19
geguileoI don't care as much about the changing of standards as of losing history14:20
hemnaanother example of sux0rs...same problem though: https://github.com/craigerl/aprsd/blob/master/aprsd/main.py#L286-L30114:20
simondodsleyeharney: then you should :)14:20
rosmaitageguileo: ++14:20
eharneyi'd rather just stick to pep8/hacking style-wise14:20
geguileoeharney: and now you've seen what it does  };-)14:20
eharneygeguileo: indeed14:21
rosmaitai don't see the advantage of using this on the main cinder code because of the history issue14:21
rosmaitamaybe for drivers? how do people feel about that?14:21
simondodsleywould it be acceptable for new code?14:21
rosmaitaor do we need to be completely consistent?14:21
geguileomaybe also for tests?14:21
whoami-rajatL#289-299 doesn't look good...14:21
sfernandgit blame is critical for new devs, specially when trying to understand critical changes between os releases14:21
e0neit doesn't make sense to have different code styles in the same repo14:21
eharneydoes black formatted code pass pep8/hacking most of the time?14:21
rosmaitasfernand: i agree14:22
hemnaI think if a driver dev wants to run it on their driver and is ok with losing history that's one thing, still sucks for other people that touch those drivers, but running it on cinder core I think would not be good.14:22
simondodsleythe style it creates make seeing code changes much simpler and makes for easier coding of changes going forward14:22
geguileosfernand: and for old devs (like me) as well ;-)14:22
simondodsleyI'm OK with loosing history for my driver code14:22
hemnaalso, backporting fixes to previous releases where black wasn't run.....sux0rs14:23
rosmaitai have been seeing some black-esque stylings in some patches recently14:23
rosmaitahemna: that is a really good point14:23
geguileohemna: right!!! backports!!!14:23
simondodsleyTrue - backports will break a lot.  OK - I'll drop my thoughts on `black` for cinder14:23
geguileoI think we agree not to do it for core code and let driver maintainers choose whether to do it or not14:24
jungleboyjhemna:  There is a huge issue.  Backports.14:24
e0newe know how to deal with backports. we have such rules for python3 only code14:24
whoami-rajatgeguileo: but only on patches that they don't want to backport?14:24
jungleboyjI would prefer people don't use it.  Definitely do not want it on core code.14:24
e0neit's time for out-of-tree drivers discussion14:24
jungleboyjWhen I did the review I thought it sucked but went down the route of, 'Well, it is their driver'14:25
geguileowhoami-rajat: well, if they want to backport them they'll have to fix the merge conflicts14:25
* jungleboyj does a hard eye roll14:25
rosmaitaok, so it's looking like the consensus is definitely no black reformatting in the main code14:25
toskydo I understand it correctly the whole issue is due to black not being configurable to be tuned to produce the same style we use today?14:25
hemnae0ne lolz14:25
rosmaitaok, we are out of time on this14:26
simondodsleyok - let's say No - I'll refactor my patch14:26
rosmaitadoes someone want to take an action item to investigate black configuration?14:26
rosmaitaso, for now ... no black14:26
rosmaitawe can revisit if someone comes up with a way to address the points made above14:26
rosmaitathanks, simon14:26
rosmaita#topic xena specs review14:27
rosmaitawell, it's spec freeze time again14:27
jungleboyjrosmaita:  ++14:27
rosmaita#link https://review.opendev.org/q/project:openstack%252Fcinder-specs+status:open14:27
rosmaitalet' see what we got14:27
hemnahrmm user visible extra specs14:28
rosmaitai think the user-visible extra specs is in good shape, a revision was just pushed14:28
hemnaI'm still not terribly sold on that 14:28
rosmaitahemna: you need to act fast and review14:28
hemnaok will do.   I don't think I'd want my customers to see my volume type extra specs.14:29
hemnait's very terse and driver specific14:29
hemnadetails that end users shouldn't see IMHO14:29
rosmaitawell, it's going to be limited to a specific subset of extra-specs keys14:29
rosmaitadefined in code, not configurable14:29
abishophemna: driver specific specs are absolutely excluded from the feature14:29
tbarronhemna: they wont see any backend specific extra specs14:29
hemnaI'll read through it14:29
whoami-rajathemna: driver specific extra specs won't be visible14:29
tbarronbackend-revealing14:30
hemnaI'm just saying it's problematic for my customers14:30
rosmaitawell, it will be governed by policy, so you can always turn it off for end-users14:30
tbarronand your customers can set policy to not show any if they want14:30
hemnawait, customers can set policy ?!  or admins?14:31
tbarronadmins14:31
hemnanm, I'll read the spec.14:31
tbarroni thought your customers were admins14:31
hemnano14:31
hemnaI'm the admin :)14:31
tbarronif you are the admin then you do it if you want14:31
rosmaitaanyway, give the spec a read, i think most of the issues have been worked out14:32
rosmaitanmve-of connection agent14:32
rosmaita#link https://review.opendev.org/c/openstack/cinder-specs/+/79636514:32
rosmaitai think the spec captures the discussion at the PTG14:32
rosmaitaso i think we're in agreement with the general direction14:33
rosmaitait will be useful for people to look it over for anything that needs to be spelled out more clearly14:33
rosmaitaSpec for "Consistent and Secure RBAC" in Cinder14:33
rosmaita#link https://review.opendev.org/c/openstack/cinder-specs/+/79765514:34
rosmaitawe already agreed to do this, but i realized that we didn't have a spec to track it14:34
rosmaitaso there it is14:34
hemnanice14:34
rosmaitaeric: did your remove force from snapshots spec merge already?14:35
hemnaI +A'd it this morning14:35
rosmaitacool14:35
eharneyyes14:35
rosmaitai was worried there was a problem with my gerrit query14:35
rosmaitaok, the rest have -1s14:36
hemnaI +A'd the temorary resource tracking spec as well14:36
rosmaitagood14:36
rosmaitayou have been busy this morning14:36
hemnayup :)14:36
rosmaitaSupport revert any snapshot14:37
rosmaita#link https://review.opendev.org/c/openstack/cinder-specs/+/73611114:37
rosmaitathere seems to be a failure of communication on this one14:37
rosmaitai asked the author to read the midcycle discussion14:38
rosmaitawhich stressed the importance of adding test scenarios to explain how this will work14:38
rosmaitabut that hasn't happened14:38
eharneyi think the spec is still missing info about safely handling operations that fail half-way through14:39
rosmaitaok, i will add a comment explaining more explicitly what needs to be addressed14:40
rosmaitait would be helpful if the proposer attended the cinder meetings so we could discuss14:41
jungleboyjrosmaita:  ++14:41
rosmaitaMigration support for a volume with replication status enabled14:41
rosmaita#link https://review.opendev.org/c/openstack/cinder-specs/+/76613014:41
rosmaitalooks like geguileo has some concerns on this one14:41
rosmaitaand it looks like the author is not willing to provide the details requested14:42
geguileorosmaita: I don't know if I'm failing to explain what's missing14:43
rosmaitaand those details are not unreasonable, they are key to understanding how this whole thing should work14:43
eharneythe main concern is about migration would interfere with the replication, right?14:43
rosmaitathis is another one that would have benefitted from discussion during the weekly meetings14:44
geguileoeharney: that is the primary, yes14:44
geguileoeharney: and iirc you also mentioned it before14:45
rosmaitaUpdate original volume az14:45
rosmaita#link https://review.opendev.org/c/openstack/cinder-specs/+/77843714:45
rosmaitathis hasn't been updated since my midcycle comment, so not looking good14:45
rosmaitaAllow volumes to be part of multiple volume groups14:47
rosmaita#link https://review.opendev.org/c/openstack/cinder-specs/+/79272214:47
rosmaitai would've sworn that we had discussed it, but maybe not14:48
geguileorosmaita: we did discuss it14:49
geguileoat least twice...14:49
rosmaitaok, i thought maybe i was dreaming about cinder again14:49
geguileolol14:50
rosmaitaok, it hasn't been updated since 24 may, so i guess the proposer has lost interest14:50
eharneyi'm concerned that this one is lacking some detail14:50
whoami-rajatrosmaita: I've left some comments on your srbac patch14:50
geguileoiirc the first time someone explained the use cases where this was useful14:50
rosmaitawhoami-rajat: ty14:50
eharneyit's not obvious what all the various things are that can happen when moving volumes in and out of consistency/replication groups14:50
eharneybut i suspect it's more than just updating markers to say it's in those groups...14:51
eharneyi.e. what happens with a chain of cgsnapshots when each snapshot has different volumes in it14:51
rosmaitaok, i'll send emails out about 736111, 766130, 778437, 79272214:54
rosmaitathe aim of the email being that they need to work with the team to address the issues14:55
rosmaitaand if it's a TZ problem for the weekly meeting, we'll have to work something out14:55
rosmaitabut i don't realistically see any of those being ready for Xena14:55
rosmaitaok, i guess that's it for specs14:56
rosmaitaplease review the ones mentioned at the beginning of this part of the meeting14:56
rosmaitathey need to merge by 23:59 UTC or thereabouts on friday14:56
rosmaitaany questions, comments?14:56
jungleboyjQuestion on the NVMe-oF14:56
jungleboyjhttps://review.opendev.org/c/openstack/cinder-specs/+/79636514:57
jungleboyjSo, There are +1s on it.  No +2s .14:57
jungleboyjWhat do we want to do about that?14:57
whoami-rajatI've left some queries and suggestions (personal) to make it more readable14:58
hemnathere is some good feedback on that review that's not been updated14:58
rosmaitai agree with Jay's comment that a more thorough os-brick person than me needs to give it a +214:59
jungleboyjhemna:  Ok.  That was why I wasn't +2.  Just trying to interpret the other +1s 14:59
hemnawe had a discussion about this agent ages ago no?14:59
hemnaI don't remember all of it15:00
rosmaitayes, at the PTG15:00
rosmaita#link https://wiki.openstack.org/wiki/CinderXenaPTGSummary#NVMe-oF_and_MDRAID_replication_approach_-_next_steps_for_connector_and_agent15:00
hemnaI'll read through it15:00
rosmaitaok, thanks15:00
rosmaitalike i said, i think it captures what we talked about15:00
rosmaitaalso, i am worried that the spdk CI has disappeared15:01
rosmaitaalthough i guess that doesn't matter for this spec15:01
rosmaitaok, we are out of time ... thanks everyone, and please complete your spec reviews as soon as you can15:01
rosmaita#endmeeting15:02
opendevmeetMeeting ended Wed Jun 23 15:02:10 2021 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)15:02
opendevmeetMinutes:        https://meetings.opendev.org/meetings/cinder/2021/cinder.2021-06-23-14.00.html15:02
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/cinder/2021/cinder.2021-06-23-14.00.txt15:02
opendevmeetLog:            https://meetings.opendev.org/meetings/cinder/2021/cinder.2021-06-23-14.00.log.html15:02
jungleboyjThanks!15:02
rosmaitathat bot can't handle topics, but it's pretty snappy for start/end meeting!15:02
whoami-rajatthanks15:02
whoami-rajat:D15:02
*** dviroel is now known as dviroel|out21:13

Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!