16:00:44 <jungleboyj> #startmeeting cinder
16:00:45 <openstack> Meeting started Wed Feb 27 16:00:44 2019 UTC and is due to finish in 60 minutes.  The chair is jungleboyj. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:48 <openstack> The meeting name has been set to 'cinder'
16:01:02 <whoami-rajat> Hi
16:01:06 <eharney> hi
16:01:13 <e0ne> hi
16:01:15 <jungleboyj> courtesy ping:  jungleboyj diablo_rojo, diablo_rojo_phon, rajinir tbarron xyang xyang1 e0ne gouthamr thingee erlon tpsilva ganso patrickeast tommylikehu eharney geguileo smcginnis lhx_ lhx__ aspiers jgriffith moshele hwalsh felipemonteiro lpetrut lseki _alastor_ whoami-rajat yikun rosmaita enriquetaso hemna _hemna
16:01:19 <geguileo> hi! o/
16:01:23 <jungleboyj> @!
16:01:23 <_pewp_> jungleboyj ＼(-_- )
16:01:24 <rosmaita> o/
16:01:27 <xyang> hi
16:01:33 <yikun> o/
16:01:36 <erlon> hola!
16:01:37 <walshh_> hi
16:01:46 <woojay> hi
16:02:13 <jungleboyj> Agenda:
16:02:17 <jungleboyj> #link https://etherpad.openstack.org/p/cinder-stein-meetings
16:02:18 <smcginnis> o/
16:02:52 <jungleboyj> Looks like we have a quorum.  So, lets get started
16:03:02 <jungleboyj> #topic Announcements
16:03:24 <jungleboyj> So, reminder that Milestone 3 is coming up next week.
16:03:35 <jungleboyj> #link https://releases.openstack.org/stein/schedule.html
16:04:11 <jungleboyj> Non-Client Libraries will be released next week.
16:04:24 <jungleboyj> #link https://review.openstack.org/#/q/project:openstack/os-brick+status:open
16:04:36 <jungleboyj> Is there anything critical in there that needs review?
16:05:17 <jungleboyj> Do we want to try to get Avishay's change in?
16:05:36 <smcginnis> It seems a little risky this late in the cycle, but wanted to see what others thought.
16:05:42 <jungleboyj> Need geguileo and hemna to weigh in there.
16:05:47 <jungleboyj> smcginnis: ++
16:05:54 <enriquetaso> o/
16:05:54 <smcginnis> Also wasn't sure about sahid's patch.
16:06:24 <geguileo> smcginnis: I am also cautious of the risks
16:06:38 <jungleboyj> smcginnis:  Yeah, that one isn't clear to me.
16:07:31 <smcginnis> Avishay's might be a good one, but it would probably be safer to wait until Train opens so we have the whole cycle to watch for unintended side effects and have time to revert or fix if needed.
16:07:50 <jungleboyj> smcginnis:  ++ That is my feeling as well.
16:07:53 <geguileo> one silly question, which CI jobs are testing os-brick master?
16:08:02 <erlon> smcginnis, +1
16:08:11 <jungleboyj> The fix could be backported if necessary.  Right?
16:08:12 <geguileo> because if most of them aren't, then we might as well review and merge it
16:08:39 <erlon> geguileo, so we can react at the time the change breaks?
16:08:47 <smcginnis> geguileo: There are jobs on the os-brick repo that run tempest with the latest os-brick from source. Is that what you mean?
16:09:05 <geguileo> smcginnis: no, I meant what Cinder CI jobs run with os-brick master
16:09:12 <erlon> we have had situations in the past that our CI broke after os-brick landed on pip
16:09:31 <geguileo> because if we don't have many, then the only chance of detecting if things break with os-brick is to merge the patch and release os-brick
16:09:51 <smcginnis> We don't have a "*-tips" job like some other projects have been doing. But we do have Cinder tests running against os-brick patches.
16:10:32 <smcginnis> geguileo: Yeah, that is a concern. But kind of why I would like to wait for train so there's not a last minute firedrill getting Stein out.
16:10:48 <jungleboyj> smcginnis:  ++
16:10:52 <erlon> smcginnis, those tips jobs suppose to run on os-brick/master?
16:11:31 <geguileo> smcginnis: but even if we merge it at the begining of train, it won't get tested in Cinder until we release os-brick in Train, right?
16:11:39 <smcginnis> erlon: From the ones I've seen, on other repos they run tempest or some subset of tests, but pull certain dependencies from source rather than latest releases.
16:12:19 <smcginnis> geguileo: Right, but we can do that at any point in Train. And even if we wait until Train-1, that gives us ~4 months to identify problems versus ~4 weeks.
16:12:28 <erlon> smcginnis, hmm, that's what we are doing for os-brick. I think that Cinder should have a few jobs doing that too
16:12:31 <geguileo> smcginnis: that's true
16:12:40 <erlon> LIBS_FROM_GIT=os-brick
16:12:56 <smcginnis> erlon: Oh right. I think we've had this discussion before. :)
16:13:04 <geguileo> erlon: that's not the only change necessary
16:13:27 <geguileo> because CI's are running with a flag to prevent cloning
16:13:28 <smcginnis> And as mentioned, we can always backport this once we are comfortable that it doesn't cause issues.
16:13:30 <erlon> geguileo, why not?
16:13:35 <geguileo> because CI's are running with a flag to prevent cloning
16:13:47 <geguileo> you'll need it to be added to PROJECTS
16:13:59 <geguileo> at least that's what I'm having to do for cinderlib
16:14:19 <smcginnis> That matches what whoami-rajat and I saw with an osc -tips job.
16:14:24 <geguileo> oh, wait... maybe for os-brick is different and that's enough
16:14:40 <jungleboyj> So, I am going to make the safe call that we don't merge this right now and then we can backport after it has had greater runtime.
16:14:45 <geguileo> anyway, I think we should change it so that Cinder jobs always run with os-brick master
16:14:54 <erlon> geguileo, hmm, I think it works for us becouse we are on zuulv2
16:14:56 <erlon> http://13.56.179.158/logs/35/639635/1/upstream-solidfire-check/cinder-solidfire-iSCSI/8174736/logs/local.conf.txt.gz
16:14:56 <geguileo> jungleboyj: +1
16:15:09 <jungleboyj> Given that avishay isn't here arguing for it.  :-)
16:15:46 <smcginnis> jungleboyj: +1
16:16:00 <erlon> geguileo, +1
16:16:07 <jungleboyj> Anything else on os-brick?
16:16:24 <smcginnis> geguileo, hemna: You're eyes on https://review.openstack.org/#/c/638639/ might be good, but I don't think that's a release critical thing either.
16:16:33 <smcginnis> *your
16:16:47 <geguileo> erlon: I think I was mistaken, iirc the code adds LIBS_FROM_GIT to PROJECTS at some point
16:17:19 <jungleboyj> Yeah, that one needs to be looked at too.  Also, not totally clear exactly what the problem being fixed is.
16:17:30 <geguileo> smcginnis: I'll have a look
16:17:42 <smcginnis> Thanks!
16:17:42 <geguileo> smcginnis: but I don't like the idea of using iscsi_id again
16:17:46 <smcginnis> Yeah
16:17:56 <geguileo> that broke like crazy in my fault injection tets
16:17:58 <geguileo> tests
16:18:10 <jungleboyj> :-(
16:19:18 <jungleboyj> Also want to note that python-cinderclient release is coming up the week after next.
16:19:39 <jungleboyj> I didn't see much that needed attention there, but wanted to make everyone aware.
16:20:25 <jungleboyj> Ok.  Finally, just a note that I have requested space for a Project Update and an On-Boarding session at the Summit.
16:20:32 <jungleboyj> So, the usual requests there.
16:20:52 <jungleboyj> That is all I had for announcements.
16:21:30 <jungleboyj> #topic Need Topics for the Denver Forum
16:21:41 <jungleboyj> #link https://etherpad.openstack.org/p/cinder-denver-forum-brainstorming
16:21:56 <smcginnis> Forum session selection is coming up soon I think, right?
16:22:32 <jungleboyj> Yes.  I believe I need to submit the topics in the next week.
16:22:39 <jungleboyj> Let me see if I can find that note.
16:23:04 <jungleboyj> Yeah, need to submit them by 3/8
16:24:26 <jungleboyj> I just added the Driver Capability Reporting topic.
16:25:06 <jungleboyj> Does anyone have other ideas that could use User/Operator input?
16:26:00 <jungleboyj> Ok.  Will make one more call next week and then submit what we have.
16:26:49 <jungleboyj> Next topic
16:26:59 <jungleboyj> #New Patches with CGs Support
16:27:06 <jungleboyj> #topic New Patches with CGs Support
16:27:18 <jungleboyj> #link https://review.openstack.org/#/c/574119/
16:27:22 <jungleboyj> e0ne:
16:27:35 <e0ne> I don't remember what was our decision on this topic
16:28:06 <e0ne> but I don't understand why do we need implement new feature for deprecated API
16:28:19 <eharney> the question is CGs vs generic groups?
16:28:54 <e0ne> eharney: that patch add CGs support without generic groups support
16:29:02 <erlon> e0ne, they might be trying to implement that in a old old OS version
16:29:52 <erlon> it should be easy to add generic groups support on that
16:30:08 <eharney> what does a driver have to actually do to support generic groups?
16:30:15 <smcginnis> Yeah, we said CGs are dead and only work on generic volume groups with consistency should be done.
16:30:15 <eharney> have to actually do*
16:30:20 <e0ne> but we're going to remove CGs at all
16:30:43 <jungleboyj> Right.  We need to go forward
16:30:48 <smcginnis> I thought we had rerouted CG APIs to call the volume group code.
16:30:55 <eharney> smcginnis: i think this is CGs via generic groups if i'm reading the code right..
16:30:59 <jungleboyj> smcginnis:  That sounds familiar.
16:30:59 <e0ne> https://review.openstack.org/#/c/574119/10/cinder/volume/drivers/dell_emc/unity/driver.py - line #59
16:31:07 <smcginnis> eharney: OK, haven't really looked.
16:31:24 <smcginnis> I wonder if xyang is around and remembers. :)
16:31:36 <eharney> e0ne: so aren't generic groups just a manager operation anyway?
16:32:00 <e0ne> eharney: it has a bit different API
16:32:03 <erlon> eharney, usually just wrap the CG functions: https://github.com/openstack/cinder/blob/master/cinder/volume/drivers/solidfire.py#L1732
16:32:13 <eharney> the manager catches NotImplementError and calls _create_group_generic..
16:33:44 <erlon> jgriffith, always making reading code fun :)
16:34:08 <eharney> i'm not a cg expert, but it looks at a glance like the driver is doing the right thing
16:34:41 <e0ne> it just confused me that driver adds CGs support
16:34:53 <jungleboyj> Oh, I miss reading jgriffith s code.
16:34:53 <eharney> it adds support for create_group()
16:35:03 <jungleboyj> Delightfully passive agressive.
16:35:05 <eharney> not create_consistencygroup()
16:35:11 <erlon> eharney, yes, it its
16:35:55 <e0ne> eharney: also, it uses 'skip_if_not_cg' decorator
16:35:58 <jungleboyj> Yeah, so it looks like it is doing the right thing.
16:36:10 <eharney> e0ne: right, because if it's not a CG, the driver doesn't have anything to do...
16:36:19 <jungleboyj> They have  the driver history explaining it that way too.
16:36:23 <erlon> though they should handle cases where the group being created is not 'consistent'
16:36:37 <eharney> erlon: they are handled
16:37:00 <erlon> eharney, ow, in the decorator, missed that
16:37:06 <e0ne> ok, looks like the patch is valid
16:37:25 <jungleboyj> :-)
16:38:05 <jungleboyj> e0ne:  Good to move on then?
16:38:10 <e0ne> jungleboyj: yes, thanks
16:38:20 <jungleboyj> e0ne:  Welcome.
16:38:38 <jungleboyj> #topic release from stable/pike->  make it 11.2.0 ?
16:38:43 <jungleboyj> rosmaita:
16:38:47 <rosmaita> hello
16:38:52 <jungleboyj> @!
16:38:53 <_pewp_> jungleboyj ＼( ･_･)
16:39:05 <rosmaita> we are about 65 or so commits behind on stable/pike cinder release
16:39:25 <rosmaita> some stuff has crept in, tony wants to make it 11.2.0 instead of 11.1.x
16:39:39 <jungleboyj> Yikes.  That was a lot of commits.
16:39:52 <rosmaita> looks like all the questionable changes (new config opts, mainly) have release notes already
16:39:56 * jungleboyj is glad that rosmaita it helping us keep up with that.
16:40:22 <rosmaita> so, need some stable cores to take a quick look and let me know whether we want to revert some stuff, or jsut release as 11.2.0
16:40:34 <jungleboyj> rosmaita:  Why do those require a version bump like that if they aren't braking existing functionality?
16:41:00 * rosmaita thinking
16:41:02 <smcginnis> We are adding config options that sometime change behavior. Plus there were some requirements changes that maybe shouldn't have been allowed.
16:41:17 <rosmaita> i also think maybe 2 opts changed default values?
16:41:27 <smcginnis> Yeah
16:41:51 <rosmaita> #link https://review.openstack.org/#/c/636401/
16:41:58 <rosmaita> see tony's comment at the bottom
16:42:00 <eharney> one driver started requiring "fallocate" which is sort of a requirements change
16:42:13 <jungleboyj> Yeah, I just read that.
16:42:31 <jungleboyj> Ok.
16:42:45 <jungleboyj> Well, I think it would be a bigger mess to try and revert things.
16:42:51 <rosmaita> i agree!
16:42:55 <jungleboyj> I am ok with bumping to 11.2.0
16:42:58 <eharney> there's no reason to revert any of this that i see, just ship it as 11.2.0
16:43:20 <rosmaita> ok, i will update the release patch
16:43:35 <e0ne> eharney: +1
16:44:00 <jungleboyj> rosmaita:  Thank you.
16:44:31 <jungleboyj> We bump the major release with each development cycle so it doesn't really impact anything.
16:45:05 <jungleboyj> So, lets move on.
16:45:08 <smcginnis> And that will better communicate to downstream consumers that there are more than just simple bugfixes included so they should pay attention to release notes.
16:45:19 <jungleboyj> #topic Cinderlib status update
16:45:21 <jungleboyj> geguileo:
16:45:25 <smcginnis> rosmaita: Commit message
16:45:32 <geguileo> quick update on cinderlib
16:45:33 <rosmaita> :)
16:45:41 <smcginnis> ;)
16:45:45 <geguileo> we now have the cinderlib project as openstack/cinderlib
16:45:52 <jungleboyj> We are back to regular updates from geguileo   :-)
16:45:54 <geguileo> and governance added it under the Cinder team
16:46:06 <geguileo> there is a patch ready for review to add CI jobs
16:46:18 <geguileo> https://review.openstack.org/#/c/638442/
16:46:40 <geguileo> ananother to add it to devstack
16:46:41 <geguileo> https://review.openstack.org/#/c/638392/
16:46:53 <whoami-rajat> the patchsets show a lot of struggle.
16:47:11 <geguileo> whoami-rajat: yeah, the ceph job was a pita
16:47:24 <geguileo> because it required the ceph-common library
16:47:30 <geguileo> that install pythong-six from rpm
16:47:47 <geguileo> but the gate installs tox from pypi, which bring six from pypi
16:47:55 * smcginnis is happy not to be the only one that typos pythong
16:48:05 <geguileo> rofl
16:48:15 <jungleboyj> pythong is my favorite version of python
16:48:16 <smcginnis> :)
16:48:25 <geguileo> and the system python site-package got into a weird state
16:48:48 <geguileo> there's also the patch for devstack-gate, but that one won't pass for the time being
16:48:52 <geguileo> https://review.openstack.org/#/c/639614/
16:49:12 <geguileo> and we have the first real patch for cinderlib, by courtesy of hemna !!!!
16:49:31 <hemna> :P
16:49:33 <whoami-rajat> so the struggle was real. glad it finally passed.
16:49:35 <jungleboyj> Yay!  Good to have hemna back!
16:49:51 <hemna> :)  it's good to be back
16:49:55 <geguileo> once the devstack code merges I'll add cinderlib functional tests to Cinder LVM and Ceph jobs as well
16:49:57 * jungleboyj waves goodbye to the struggle bus
16:50:05 <geguileo> that's the update
16:50:41 <geguileo> oh, and once the CI jobs are merged
16:50:42 <jungleboyj> geguileo:  Sounds good.  Thank you for being patient and working to move this to its own project.
16:51:22 <jungleboyj> So, the final topic.
16:51:28 <geguileo> jungleboyj: thank you all for steering this into the right direction
16:51:38 <jungleboyj> #topic Feature Freeze Doubts
16:51:45 <jungleboyj> whoami-rajat:
16:52:11 <jungleboyj> So, untyped volume spec should go into Train.
16:52:21 <jungleboyj> Too late for that to go into Stein.
16:52:52 <jungleboyj> We should also take a look and see if there are other Specs that merged but don't have associated code merged.
16:53:23 <jungleboyj> As for the api-ref automation ....
16:53:25 <whoami-rajat> jungleboyj:  ok. The code also won't be ready by the freeze time so good for me :D Thanks
16:53:26 <jungleboyj> Hmmm,
16:53:58 <jungleboyj> I think those are still ok to merge until RC unless anyone disagrees.
16:54:15 <smcginnis> ++
16:54:25 <yikun> +1
16:54:51 <smcginnis> yikun: Staying up late again I see. :)
16:55:01 <yikun> :)
16:55:09 <jungleboyj> :-)
16:55:22 <jungleboyj> 😴
16:55:32 <whoami-rajat> jungleboyj: so can still work on the api-ref responses after freeze ? right?
16:55:52 <jungleboyj> Correct.  Have a few more weeks until we hit the RC point.
16:56:24 <whoami-rajat> jungleboyj: ok. Thanks for the clarifying my doubts. :)
16:56:32 <jungleboyj> whoami-rajat:  No problem.
16:56:52 <jungleboyj> whoami-rajat: Happy to help new people learn the process.
16:57:02 <jungleboyj> It is a good thing.
16:57:10 <jungleboyj> #topic Open Discussion
16:57:23 <jungleboyj> Anything else we need to talk about or can we end on time for once?  :-)
16:58:01 <eharney> i found that a bunch of CI jobs are running too many tempest scenarios, trying to get those nice and cleaned up
16:58:13 <eharney> should improve failure rate for ceph and lio-barbican
16:58:15 <eharney> just an fyi
16:58:32 <jungleboyj> eharney:  Nice.  Would like to see those failing less often!
16:59:10 <jungleboyj> Lets wrap up then.
16:59:23 <jungleboyj> One last reminder to please give input to the Forum planning.
16:59:37 <jungleboyj> Thank you for taking time to attend the meeting and to work on Cinder
16:59:44 <jungleboyj> Have a great week everyone.
16:59:50 <whoami-rajat> jungleboyj: Thanks!
16:59:52 <jungleboyj> #endmeeting