16:00:31 <jungleboyj> #startmeeting cinder
16:00:32 <openstack> Meeting started Wed Nov 29 16:00:31 2017 UTC and is due to finish in 60 minutes.  The chair is jungleboyj. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:33 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:36 <openstack> The meeting name has been set to 'cinder'
16:00:41 <Swanson> Hi
16:00:45 <smcginnis> o/
16:00:49 <jungleboyj> courtesy ping:  jungleboyj DuncanT diablo_rojo, diablo_rojo_phon, rajinir tbarron xyang xyang1 e0ne gouthamr thingee erlon patrickeast tommylikehu eharney geguileo smcginnis lhx_ lhx__ aspiers jgriffith moshele hwalsh
16:00:57 <tommylikehu> hi
16:00:57 <gouthamr> hi o/
16:01:00 <rajinir> o/
16:01:02 <erlon> hey
16:01:52 <jungleboyj> Hey all.
16:02:49 <jungleboyj> Ok, well, looks like this is the group for today.  Lets get started
16:02:57 <jungleboyj> #Announcements
16:02:59 <abishop> o/
16:03:34 <jungleboyj> Please continue to watch review priorities.  There have been a couple of specs that merged this week.  Thank you for the help with reviews there.
16:03:50 <jungleboyj> #link https://review.openstack.org/#/c/523608/
16:04:10 <jungleboyj> That is the multi-attach spec.  We should get attention to that so we can hopefully get that merged.
16:04:29 <jungleboyj> The Inspur driver merged.
16:04:42 <jungleboyj> Still waiting for DataCore and Hedvig to work out CI.
16:04:53 <jungleboyj> Appreciate cores keeping an eye on those in the next week.
16:05:24 <jungleboyj> The usual reminder that Milestone 2 is next week.
16:06:03 <jungleboyj> smcginnis:  When would you normally cut the milestone ?
16:06:47 <smcginnis> jungleboyj: I would try to shoot for the Wednesday of the deadline, so it's not a rush on Thursday.
16:06:54 <smcginnis> Thursday is the actual deadline.
16:07:10 <jungleboyj> Ok, that was what I thought.  So we have 1 week to get drivers/specs in.
16:07:19 <jungleboyj> smcginnis:  Thanks for verifying that.
16:07:52 <jungleboyj> The PTG dates have been announced officially!
16:08:08 <jungleboyj> #link https://www.openstack.org/ptg/
16:08:17 <jungleboyj> Croke Park, Dublin, Ireland 2/26 to 3/2/2018
16:08:33 <jungleboyj> How many people here think they will be able to attend?
16:08:36 <erlon> jungleboyj: what are the expected cuts (features, drivers, etc )
16:08:52 <jungleboyj> erlon: Specs Freeze and Driver freeze
16:09:23 <erlon> 2 -> spec freeze 1: Driver freeze?
16:09:23 <geguileo> jungleboyj: I think I'll be going  :-)
16:09:36 <jungleboyj> I have sent a note indicating that Cinder will participate and estimated 15 people coming.
16:09:53 <tommylikehu> 15!
16:10:07 <jungleboyj> I am in, I assume smcginnis is in, Helen and maybe some of her co-workers.
16:10:11 <Swanson> That a lot?
16:10:21 <smcginnis> jungleboyj: I'll probably be there. :)
16:10:21 <jgriffith> Swanson: no
16:10:32 <jungleboyj> tommylikehu:  Yeah, that is small.
16:10:36 <jungleboyj> jgriffith:  You in?
16:10:43 <jgriffith> jungleboyj: planning on it
16:10:51 <smcginnis> Maybe DuncanT can stop by.
16:10:57 <jgriffith> LOL
16:11:07 <jgriffith> he better, or we all show up at his door an harass him
16:11:17 <jungleboyj> jgriffith: Awesome.
16:11:29 <jungleboyj> DuncanT:  That will be awesome to see him again!
16:11:55 * jungleboyj looks forward to jgriffith and DuncanT disagreeing on everything again.  ;-)
16:12:06 <Swanson> 1 person per 10 drivers?
16:12:47 <jgriffith> Swanson: better ratio than reviewers :)
16:12:54 * DuncanT will be in Dublin
16:13:06 <smcginnis> jgriffith: So true. Sadly.
16:13:09 <jungleboyj> Anyway, please start planning on attending if possible.  I will take a poll as we get closer
16:13:13 <jungleboyj> smcginnis:  :-(
16:13:23 * jungleboyj celebrates DuncanT 's response!
16:13:52 <jungleboyj> Ok ... Oh and one other announcement.
16:14:03 <jungleboyj> I got python-brick-cinderclient-ext 0.6.0 yesterday.
16:14:22 <jungleboyj> It was broken by removal of the connector constants from os-brick .
16:14:30 <jungleboyj> That will be fixed with the new release.
16:15:06 <jungleboyj> Moving on.
16:15:18 <jungleboyj> #topic Documentation with new features:
16:15:45 <jungleboyj> A reminder to everyone that we really need to be including documentation with patches that introduce new features.
16:16:00 <jungleboyj> smcginnis:  and I have not been blocking on those in general.
16:16:10 <jungleboyj> Will start doing that in the future though.
16:16:40 <jungleboyj> If the initial patch doesn't include documentation we need to be following up and making sure that it does get in.
16:17:00 <jungleboyj> Any questions on that item?
16:17:40 <jungleboyj> Next ...
16:18:01 <jungleboyj> #topic Oversubscription (provisioning improvements):
16:18:08 <jungleboyj> gouthamr: and erlon
16:18:15 <gouthamr> thanks jungleboyj
16:18:17 <jungleboyj> #LINK: https://review.openstack.org/#/c/490116/
16:18:21 <erlon> hey
16:18:23 <gouthamr> a'rite this is wrt
16:18:23 <gouthamr> #LINK: https://review.openstack.org/#/c/490116
16:18:26 <jungleboyj> #ling: https://review.openstack.org/#/c/490116/
16:18:33 <jungleboyj> #link: https://review.openstack.org/#/c/490116/
16:18:50 <gouthamr> :P
16:18:52 <jungleboyj> One of those will work.  :-)
16:19:01 <smcginnis> :D
16:19:27 <gouthamr> erlon/I have been addressing some user expectations downstream about this feature and we think this spec will help some of these concerns
16:20:10 <gouthamr> geguileo compiled this spec and noted 2-3 action items, however there is no assignee
16:20:36 <gouthamr> we were wondering if there was anyone working on it, and if we can help drive this forward..
16:20:38 <erlon> gouthamr: also there were some points we discussed about adding better testing for it
16:20:41 <jungleboyj> We merged the Spec yesterday.
16:21:35 <jungleboyj> geguileo: Is anyone working on the items in the spec?
16:21:48 <geguileo> jungleboyj: no
16:21:56 <jungleboyj> geguileo:  Okie Dokie.
16:22:09 <geguileo> My priority when writing it was to make sure that the terms and what we have right now was in there
16:22:20 <jungleboyj> geguileo:  ++
16:22:21 <geguileo> The improvements were a bonus  XD
16:23:08 <gouthamr> nice, thanks for that geguileo.. we do think the automatic calculation will help some of these users who can't understand the behavior of this inside the scheduler
16:23:12 <jungleboyj> geguileo:  That is always a good thing.
16:23:15 <erlon> geguileo: Nice, that was really good to have those defined
16:23:24 <jungleboyj> erlon: ++
16:24:06 <erlon> jungleboyj: we should add those to the devref as well
16:24:19 <gouthamr> +1
16:24:23 <jungleboyj> erlon: ++
16:25:05 <jungleboyj> So, gouthamr and erlon  are you guys volunteering to help implement things from the spec?
16:25:17 <gouthamr> yessir. if anyone would like to help out, please get in touch with erlon and gouthamr, we'll put together a patch..
16:25:43 <jungleboyj> gouthamr: Excellent
16:26:10 <smcginnis> Thanks for taking that.
16:26:22 <jungleboyj> #action gouthamr and erlon to put together a patch to implement geguileo 's spec.
16:26:44 <smcginnis> It did concern me that the spec has "Primary assignee: None" :)
16:26:46 <jungleboyj> erlon: Would you be willing to make sure the information gets into the dev ref?
16:27:06 <jungleboyj> smcginnis:  Agreed but it did also indicate that the primary point of the spec was documentation.
16:27:23 <erlon> jungleboyj: what do you mean?
16:27:54 <smcginnis> Yes, true. But documentation that points out current issues that needs someone to fix.
16:28:02 <jungleboyj> :-)  True.
16:29:01 <jungleboyj> erlon:  You noted that the definitions from the spec and algorithms should make it into the devref.  Will you just make sure that is part of your patch or is that additional work?
16:29:18 <gouthamr> there were two proposals: one was on automatically computing the max_over_subscription_ratio inside the scheduler
16:29:33 <erlon> jungleboyj: Ow, ok, I can do that, just need to compile a doc from the spec
16:29:55 <erlon> but should go in a separate patch I assume
16:30:01 <gouthamr> and the other was to report bugs against errant drivers - this one would need help from other driver maintainers..
16:30:07 <jungleboyj> erlon: :-)  That was what I was getting at.
16:30:53 <jungleboyj> I mean, we could maybe just point to the spec, but I think having a more official document come out of the spec would be best.
16:31:34 <erlon> jungleboyj: there are some good development references in doc/source/contributor/, I think it should do there as well
16:31:48 <jungleboyj> erlon: Perfect.
16:32:28 <erlon> jungleboyj: just a question, how long until feature freeze from today?
16:32:48 <jungleboyj> erlon:  Good question.
16:32:52 <jungleboyj> #link https://releases.openstack.org/queens/schedule.html
16:33:14 <jungleboyj> Jan 19th is the Feature freeze.
16:33:15 <erlon> jungleboyj: damn I was trying not having to do the math
16:33:24 <jungleboyj> So, about a month and a half.
16:33:29 <erlon> jungleboyj: ok
16:33:35 <jgriffith> this new scheduling is annoying
16:33:35 <erlon> jungleboyj: thanks
16:33:43 <jungleboyj> erlon:  No problem.
16:33:46 <smcginnis> jgriffith: What's new?
16:33:53 <jungleboyj> jgriffith:  Yeah ... agreed.
16:34:02 <smcginnis> PTG timing within the cycle?
16:34:04 <jungleboyj> smcginnis: Used to be more or less done at Christmas.
16:34:13 <smcginnis> Ah. Yeah, that was nice.
16:34:19 <jungleboyj> smcginnis:  Yep.
16:34:31 <jgriffith> smcginnis: yeah, the whole rework of things has kinda backfired if you ask me.  Creates mass confusion
16:34:43 <jgriffith> anyway... carry on
16:34:52 <jungleboyj> jgriffith: Agreed, but c'est le vie
16:35:04 <jgriffith> jungleboyj: yep yep yep
16:35:11 <jungleboyj> gouthamr:  Ok, so I think we have an open question from you.
16:35:26 <gouthamr> do we?
16:35:30 <jgriffith> https://www.youtube.com/watch?v=tnf7r6lUVfA
16:35:47 <jungleboyj> jgriffith:  Nice.
16:35:59 <jungleboyj> gouthamr:  Well, I guess I need to better understand the plan based on the spec.
16:36:38 <jungleboyj> There is max_over_subscription_ratio calculation in the scheduler and then filing bugs against broken drivers.
16:36:41 <smcginnis> jgriffith: You want confusion - there are more people arguing for once a year releases. And others arguing for more frequent releases.
16:36:58 * jungleboyj shakes my head
16:37:29 <gouthamr> jungleboyj: yes... we can eyeball the other drivers that support this to see if they are compliant with respect to the space reporting
16:37:37 <jgriffith> smcginnis: as long as we have 12 or so deadlines within each that would be great :)
16:37:56 <jungleboyj> gouthamr:  Ok, and for those that are not will you switch them to using the in scheduler reporting or just open bugs against them.
16:38:50 <gouthamr> jungleboyj: yes.. there's a new configuration option proposed which will default to not doing the automatic calculation for backwards compatibility
16:39:42 <jungleboyj> gouthamr: Ok.  (Sorry if they are dumb questions, it was a while ago that I read the spec in detail).
16:40:12 <jungleboyj> gouthamr: So, drivers that are not doing things properly should be fixed but we can work around it with a config option.
16:41:07 <jungleboyj> That sounds good to me.
16:41:26 <erlon> jungleboyj: actually the drivers that are not doing things properly (reportin provisioned_capacity_gb), if not fixed will continue behaving in the worng way
16:41:45 <jungleboyj> erlon:  Ok, that was what I was trying to get at.
16:41:54 <erlon> jungleboyj: not sure what to do to enforce a common behaviour across all drivers
16:42:02 <jungleboyj> They would need to fix it or be changed to use the scheduler?
16:42:11 <gouthamr> jungleboyj: yes, they may be reporting incorrect values today, and that would affect the oversubscription calculations today.. we'll now add this toggle to allow administrators to let cinder adjust the max_over_subscription_ratio based off space realities..
16:42:50 <gouthamr> so these errant drivers will probably be wrong even with this optional automatic calculation
16:43:05 <erlon> gouthamr: exact
16:43:11 <jungleboyj> gouthamr:  Ok.
16:44:49 <jungleboyj> So, lets get the scheduler improvement in place, open bugs for the broken drivers and then figure out how to tackle those drivers once we know what that list looks like.
16:44:54 <jgriffith> jungleboyj: got time to talk about multi-attach spec?
16:45:00 <gouthamr> jungleboyj: +1
16:45:06 <jungleboyj> gouthamr:  Cool.
16:45:13 <jungleboyj> jgriffith:  I think we will.
16:45:18 <jgriffith> cool
16:45:22 <erlon> jungleboyj: +1
16:45:33 <jungleboyj> jgriffith: Lets wrap this up and then I have one short time and then we got you.
16:45:42 <jungleboyj> gouthamr: Anything else on this?
16:45:46 <gouthamr> jungleboyj: nope
16:46:22 <erlon> jungleboyj: Im fine
16:47:03 <jungleboyj> #action gouthamr and erlon  to put together patches for the scheduler and documentation.
16:47:19 <jungleboyj> #action gouthamr and erlon to open bugs against broken drivers
16:47:35 <jungleboyj> #action to readdress the plan for broken drivers once we know what the list looks like.
16:47:58 <jungleboyj> gouthamr: erlon  Thank you so much for picking this up.  geguileo Thanks for getting it started.
16:48:07 <geguileo> np
16:48:11 <gouthamr> +1
16:48:17 <jungleboyj> #topic Install Documentation:
16:48:31 <jungleboyj> This is a follow-up to the discussion I started last week.
16:49:01 <jungleboyj> Haged Finley has been looking at our install documentation with fresh eyes.
16:49:12 <jungleboyj> *Hagen  (datasundae) on IRC.
16:49:46 <jungleboyj> Finding a number of issues.  All of this work also falls into place with the new Contributor Portal and the updates to OpenStack Upstream Institute.
16:49:58 <jungleboyj> #link https://etherpad.openstack.org/p/cinder-documentation-redesign
16:50:18 <jungleboyj> Given that, I have put together an etherpad with some design thoughts.
16:50:38 <jungleboyj> Lot of work to be done there but I wanted to share this with the team.  Please feel free to add thoughts/concerns in there.
16:51:07 <jungleboyj> Want to make our docs easier for Users, New Contributors and experienced developers to utilize.
16:51:35 <jungleboyj> Need to fix the copy pasta mess that came from pulling it over from the the docs repo.
16:51:42 <jungleboyj> So, take a look if you have a chance.
16:51:48 <jungleboyj> Any questions there?
16:51:59 <smcginnis> Good plan.
16:52:29 <jungleboyj> smcginnis:  Thank you.  Just need time to implement.  Thankfully datasundae is making nice incremental progress.  Should see some patches here soon.
16:52:55 <jungleboyj> #topic Multi-attach spec
16:53:02 <jgriffith> cool
16:53:08 <jungleboyj> #link https://review.openstack.org/#/c/523608/h
16:53:10 <jgriffith> https://review.openstack.org/#/c/523608/2
16:53:14 <jgriffith> jungleboyj: :)
16:53:31 <jgriffith> alright, so I just wanted to point this out and mostly get input on the volume-type thing
16:53:36 <jungleboyj> Great minds ...
16:53:57 <jgriffith> My opinion has been that we shouldn't require prior knowledge, but I added the volume-type requirement to the spec anyway
16:54:00 <jgriffith> because it came up
16:54:33 <smcginnis> jgriffith: I do like that with volume types, a user could just retype if they didn't have multiattach specified.
16:54:36 <jgriffith> The dilema is unexpected behaviors for an end user if the backend doesn't support multi-attach
16:54:51 * jungleboyj is looking.
16:54:56 <smcginnis> I didn't really like the idea before that it needed to be specified at creation and never change after that.
16:55:04 <jgriffith> smcginnis: +1
16:55:11 <jgriffith> smcginnis: I saw the volume-type as a compromise
16:55:39 <jgriffith> If anybody has strong feelings let me know either via IRC or via the spec
16:55:41 <smcginnis> I think it's good. Especially if they are on a multiattach capapable backend, retype would be pretty simple.
16:56:11 <jgriffith> smcginnis: cool, that's settled then :)
16:56:16 <smcginnis> Hah
16:56:24 <jgriffith> eharney: I'll add your use case momentarily
16:56:57 <jgriffith> have to fix my pep8 errors again anyway... stupid vimrc
16:56:59 <jungleboyj> jgriffith:  Yeah, I think that makes sense.  I will read the spec after the meeting.
16:57:10 <jgriffith> jungleboyj: thanks, that's all I've got
16:57:25 <jungleboyj> Cool.
16:57:34 <jungleboyj> Team, please looks so we can also try to get that spec in.
16:57:56 <jungleboyj> jgriffith:  Thank you for all your work there.
16:58:16 <smcginnis> Thanks ildikov for pressuring jgriffith to do that work. :)
16:58:22 <jgriffith> haha!
16:58:25 <jgriffith> indeed!
16:58:36 <jungleboyj> jgriffith:  Also, FYI, it appears there are some bugs in the new API use from python-brick-cinderclient-ext .
16:58:44 <jungleboyj> Looking into those and see if I can fix them.
16:58:53 <jungleboyj> smcginnis:  +1000
16:59:32 <jungleboyj> Well, that is about time.
16:59:49 <jungleboyj> Thank you team for your work on Cinder and for coming to the meeting.
16:59:54 <jgriffith> jungleboyj: I don't imagine the ext has any implementation for the new API calls but I haven't looked
17:00:00 <jungleboyj> Have a great week and talk to you next week.
17:00:06 <jungleboyj> jgriffith:  It actually does.  :-)
17:00:13 <smcginnis> Thanks.
17:00:15 <jungleboyj> It mostly works.  :-)
17:00:33 <jungleboyj> #endmeeting