18:00:44 #startmeeting Keystone 18:00:45 Meeting started Tue Oct 14 18:00:44 2014 UTC and is due to finish in 60 minutes. The chair is morganfainberg. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:47 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 18:00:48 o/ 18:00:50 The meeting name has been set to 'keystone' 18:00:51 he 18:00:56 hi too 18:00:59 good morning! 18:01:19 o/ 18:01:20 😎 18:01:28 o/ 18:01:41 so have a couple house keeping things to look at regarding the spec process then we're on to the important thing(s): Summit planning and session finalization 18:01:43 ayoung look scary 18:01:54 😫 18:01:59 hello 18:02:01 how's that? 18:02:02 o/ 18:02:06 #topic Kilo Spec Freeze deadline and adopting spec guidelines 18:02:09 * gyee runs 18:02:27 Other projects have spec freeze deadlines. I just want to formalize it for us this cycle. 18:02:54 is K1 a reasonable target to freeze acceptance of specs for Kilo? 18:03:00 1/2 way to k2? 18:03:15 K1 feels too early to me 18:03:21 K2 18:03:29 anything that isn't accepted beyond that point would need an exception granted. -- this does mean we *need* to step up and review 18:03:38 k1 follows very quickly after the summit 18:03:53 first couple weeks in december, IIRC? 18:03:57 ayoung, i don't want the spec freeze deadline to be too late, i want to try and front load as many new features as possible this cycle to the start of development 18:04:19 yeah...with US Thanksgiving and Xmas prep messing up the milestone 18:04:25 usually ppl put up POC code along with the spec anyway so we have something to look at 18:04:31 ayoung, i was thinking either K1 or 1/2 way thorugh K2 for spec freeze (gives us a window to make sure any exceptions can be accepted by k2) 18:04:42 dolphm, fair point, K1 is too soon 18:04:42 what are other projects doing? 18:05:02 and what benefit do they get out of a spec approval deadline? 18:05:07 Nova has a cut spec freeze date, not sure what it is though 18:05:18 I’d propose m2….but remembering that getting the spec into m2 doesn’t gurantee you’re patch will make the release….just that NOT having the spec by m2 gurantees it will NOT. 18:05:22 dolphm, it limits the scope of when new code lands. 18:05:32 henrynash, agreed 18:05:34 dolphm, so we don't have tons and tons of stuff going in late in the release 18:05:43 henrynash, exactly 18:05:56 it's just making sure we keep enough time for changes to land post spec approval. 18:06:07 i'm fine with M2 if that is the general consensus. 18:06:14 changes to core APIs can also be K2 if we want to be conservative 18:06:19 when is the midterm meetup? spec usually get approve by then right? 18:06:36 gyee, mid cycle has been a couple week prior to m2 18:06:37 if we say the spec cut of is milestone 2, do we require that you have code up for review by then too? 18:06:43 I know we were shifting the emphasis to the spec reviews, but I think its OK to say "any API changes after K2 will be in extensions only" 18:06:44 lbragstad, no, just the spec approved 18:06:49 morganfainberq: and I think it puts the pressur on a proposer to try and get it into m1 since that gives them a much greater change of getting the patch by m3 18:06:49 ok 18:06:52 morganfainberg, so m2 sound pretty reasonable 18:07:01 henrynash, ++ 18:07:27 ok so lets set the deadline to m2 for approval. Exceptions can/will be granted as needed. but lets front load what we can 18:07:31 does this mean we'll be more strict about only merging patches that are part of approved specs? 18:07:42 dstanek, that is the next bit here. 18:08:01 so. another thing i like that nova is doing is trival BPs 18:08:03 No specs for trivial blueprints, just create the BP in LaunchPad and New, targeted, unapproved BPs will be reviewed in weekly meetings, If it is agreed they are indeed trivial in the meeting, they will be approved. (same as Nova), this should be minimal overhead. 18:08:26 if it isn't trivial, we force a spec 18:08:42 what's the difference between a trivial bp and a wishlist bug? 18:08:58 bknudson, i'd say same general level. 18:09:12 bknudson, i think that's morganfainberg's call to make :P 18:09:45 morganfainberq: I like it…we just have to make sure we DO discuss in the meeting, so that we don’t delay asking the proposer to go write a full spec 18:09:46 i'd rather see things as BPs than wishlist bugs. because "wishlist" isn't really a bug. 18:09:47 anything that take more than two paragraphs to explain :) 18:09:54 henrynash, correct, it'll be the first thing on the meeting each week 18:10:04 wishlist may not be trivial 18:10:25 sure. and therefore it would need a spec, just the same as a bigger bp 18:10:40 yeah, wishlist seems to indicate importance/priority, not scope 18:10:48 morganfainberg, can we have an exemption process for specs that were not proposed before k2? 18:10:48 ok. 18:11:07 stevemar, absolutely, it'll be a message to the mailing asking for an exception. 18:11:22 stevemar, then the team will either yes exception / no sorry 18:11:24 morganfainberg, iirc there were a few that were proposed late J2 or early J3 this time around 18:11:24 OK 18:11:28 lbragstad, why would you need code at time of spec? 18:12:08 henrynash, this also means any BPs that are untargeted / unapproved / etc I'm going to go cleanup in LP [if we like the trivial bp model] 18:12:13 we've only required code in the past if you're asking for an exemption, as proof that the exemption isn't in vain 18:12:21 topol: I was thinking if the spec were to be approved late in the process, it would speed up the review of the implementation 18:12:25 morganfainberg: ++ 18:12:27 dolphm, i think that is a fair request to maintain 18:12:49 cleaning up lp bps would be great. 18:13:04 people keep asking me when an old bp is going to be implemented. 18:13:08 +1 on cleaning up bps 18:13:12 ++ 18:13:14 bknudson, i admit, that was the other thing i felt we'd get out of this since we managed to do a good job on the bugs, time to solve the BPs 18:13:26 bknudson: I get those same questions... 18:13:30 great job on keeping the bugs triaged, btw 18:13:40 everyone, thanks for keeping eyes on those 18:13:41 close all the BPs, then reopen the ones we want 18:13:48 ayoung, thats the general idea 18:14:04 i might be a little more tactical if i know we have one that should be reviewed now. 18:14:28 bknudson: maybe all old BP's should be updated with some status and message saying it's blocked by a lack of approved spec? 18:14:34 i'll write up an email explaining the guidelines and we'll get a fix to the spec repo readme to show the new bp/deadline guideline. 18:15:13 dolphm, not sure if LP has such a "setting" 18:15:13 ++ 18:15:19 or maybe it's "blocked"? 18:15:34 morganfainberg: no, that raises ttx's alarm bells 18:15:43 dolphm, lets just close out the BPs that are old. 18:16:01 Definition: Drafting? 18:16:08 eh, maybe? 18:16:09 morganfainberg: but what about new ones that are opened without specs? 18:16:24 Priority: Not? 18:16:31 dolphm, i was thinking of closing out the ancient ones, any new ones opened w/o specs we'd keep (a bit more "tactical" than close them all) 18:16:48 then we can review them and say "yes, do it" or "go spec write" 18:17:00 i'll see how nova classifies them in the latter category 18:17:35 morganfainberg: that conflicts with the existing spec process, IMO, where the first part of the spec is an argument that the problem statement is actually a problem 18:17:43 theres only 141 bps, shouldn't be an issue at all 18:17:58 stevemar: what's the oldest bp? 18:18:02 stevemar: and there's no (documented?) blueprints API on LP 18:18:36 morganfainberg: Implementation: Deferred (.. pending spec)? 18:18:56 dolphm, i think my point is we should keep the spec process light for the things like "add pycadf notifications", regardless of the classifier 18:19:02 dolphm, yeah that one sounds right. 18:20:00 dolphm, i like deferred as a choice... 18:20:29 i also feel like a lot of the bps will be closed since they just haven't ever been cleaned up really 18:21:16 dolphm, or "pending approval"? 18:21:18 bknudson, i can't find a way to sort it by age 18:21:34 morganfainberg: pending approval is the default already 18:21:43 dolphm, "new" is the default 18:22:05 morganfainberg: oh, Direction: Needs Approval is what i was thinking of 18:22:09 ah 18:22:11 yeah 18:22:13 morganfainberg: i don't think we use that field either 18:22:31 no we don't let me see which fields only the drivers can set 18:22:50 we'll pick one of those (if possible) to classify needing approval / spec 18:23:04 i'll figure that out post meeting. 18:24:02 ok... so no opposition to the deadline or the trivial bp change? 18:24:07 or concerns? 18:24:08 morganfainberg, is the blueprint spec template changing any? 18:24:14 morganfainberg, again, please let me clean up my own. I have quite a few there that have been "parking lot" items 18:24:19 https://blueprints.launchpad.net/keystone/+spec/failing-tests for example 18:24:22 ayoung, please clean those up today then? 18:24:25 or tomorrow 18:24:52 Will do. Just don't want them disappearing on me. 18:24:57 ayoung, or i can classify all of them as "needing spec" and we can just revisit 18:25:00 ayoung: https://blueprints.launchpad.net/keystone/+spec/alembic https://blueprints.launchpad.net/keystone/+spec/basic-auth https://blueprints.launchpad.net/keystone/+spec/delegation-workplans https://blueprints.launchpad.net/keystone/+spec/distributed-signing 18:25:17 ah, yeah, alembic... 18:25:17 ayoung, i wont close yours out. 18:25:25 ayoung: actually, here's all 42 https://blueprints.launchpad.net/~ayoung 18:25:36 that leaves 100 18:25:47 99 18:25:58 ayoung just needs to get 2 bps done a week. 18:26:00 Heh, thought for sure I'd have more than half of them 18:26:02 ayoung, so, can i just classify all of yours as "needs" spec and we'll revisit them as you have achance to look them over? 18:26:07 i'll go through my outstanding ones after this meeting 18:26:13 morganfainberg, Im gonna work on a CADF BP for gyee, Do I use the same spec template as last time? 18:26:14 couple fo them have specs as well, but I can set those myself, so sure 18:26:21 er, have specs already 18:26:23 ayoung, if they have a spec i'll leave them 18:26:55 topol, is it something that really needs a spec in light of "trivial bp" guidelines? 18:26:57 ssl-everywhere is, I think, beyond just Keystone now. 18:27:08 topol, i'd say the previous cadf spec would have fallen into that category 18:27:16 there's ssl config in devstack now 18:27:26 morganfainberg, folks like to know what fields are gonna be there and why. I could make it fit in the trivial BP I guess 18:27:39 topol, if you feel better with the spec, use the spec :) 18:27:47 morganfinberg, maybe I just like to hurt myself :-) 18:27:57 topol, i wont say "no" to having a more formal BP. and yes the previous template you used is fine imo 18:28:02 s/template/form 18:28:07 morganfainberg, I think I try the lighter approach this time 18:28:08 topol, morganfainberg, thanks, CADF is useful for end-to-end trace 18:28:39 great. 18:28:46 is it just projects that need cadf? why add one at a time? 18:29:03 bknudson, good point! 18:29:29 bknudson, yeah, openstack services track projects so that's why its more urgent 18:29:44 projects represents ownership 18:29:58 bknudson, i was actually going to propose that we make all the 'normal' notifications into CADF ones, but was going to wait til summit 18:30:13 but I agree, it should be available for all APIs 18:30:13 oops, stole stevemar's thunder. 18:30:20 bknudson, lol 18:30:22 stevemar, ++ 18:30:31 i think dolphm gave me a todo to dig into the notification as a contract, but i never did :) 18:30:32 stevemar, good idea 18:30:51 stevemar ++ 18:30:52 why have two notifications formats 18:30:58 ayoung, ++++++++++ 18:31:01 CADF for the People! 18:31:02 stevemar: yeah, breaking consumers of existing notifications would be my only concern 18:31:06 everyone +++++ 18:31:07 and cadf provides a nice "known" format 18:31:11 but CADF for everything makes a lot of sense 18:31:31 nkinder, yeah, we would deprecate it over 2 releases of course! 18:31:33 stevemar, maybe a toggle "non-cadf compat notify" 18:31:43 ohh nice 18:31:44 with a fixed deprecation/removal timeline 18:31:56 or make it configurable of we have concerns over the interface 18:31:58 morganfainberg very cool 18:32:25 ok this can be followed up in spec/irc/etc, sounds like we totally should do cadf all the things 18:32:30 gyee: I'd prefer that we let the old one die instead of having something that is configurable long-term 18:32:50 morganfainberg, i can toss up a spec 18:32:55 stevemar, perfect! 18:33:10 nkinder, sure no argument here 18:33:40 ok so the next big topic... Summit session planning 18:33:49 #topic Kilo Summit Planning Design Session Schedule 18:33:55 #link https://etherpad.openstack.org/p/kilo-keystone-summit-topics 18:34:03 I've put up a tentative schedule of the sessions 18:34:27 we have one open slot still to fill. I'm open to merging / dropping / changing the sessions 18:34:30 and the timeslots 18:34:45 Jamie wanted to discuss policy future direction 18:34:52 and I would like to get some volunteers to lead the session(s) 18:35:03 I'm guessing it's in the same vein as the thread morganfainberg started 18:35:13 morganfainberg, i thought the session schedule was pretty good imo 18:35:28 not_jamielennox, ? 18:35:36 is not_jamielennox ever coming back? 18:35:43 bknudson, ask his wife 18:35:45 ++ for policy future discussions 18:35:47 next week I believe 18:35:51 so two *really* important things: we need the starting questiosn for the Ops session (similar to what we had laast summit) and we need to fill the last session slot 18:36:05 morganfainberg, make two slots for federation 18:36:11 policy is a good topic for discussion at summit since we can get more ops folks there. 18:36:13 so is policy worth a session, or should it be somethign more informal in our open/free discussion time? 18:36:18 K2K, saml for Horizon will both take a while 18:36:20 I'm sure they have their complaints if they ever tried to change it. 18:36:29 what about policy? 18:36:29 ...or read it 18:36:30 ayoung: ++ 18:36:38 I'd be willing to dedicate one of our sessions to the policy discussion (thread I started) 18:36:43 or another federation session 18:36:58 Better Policy Model and Representing Capabilites? 18:36:58 how about whiskey 18:37:03 I think both are important, but federation seems more immediate 18:37:05 ayoung, yes 18:37:09 not_jamielennox: ++ 18:37:14 dolphm, oh yeah :) 18:37:15 someone bring plastic cups this time 18:37:16 dolphm, that is probably going to be replaced by Wine, as we are going to be in Paris 18:37:31 the options I see for the last slot: VO, Split ID, 2nd fEderation, or Policy 18:37:41 Keystone Wine Tasting Session # it'll be the most popular session on the schedule 18:37:46 dolphm, ++ 18:37:53 morganfainberg, so after my customer visit I got more feedback on how openstack documentation needs significnat improvement. Any way to solicit feedback on where it needs to be improved? 18:37:57 hell yeah! 18:37:58 Federated Wine Tasting Policies 18:38:03 bring your wine, we'll taste it? 18:38:04 camoflauge 18:38:15 i think it's a bit early to talk about VO 18:38:27 stevemar: VO -> virtual organizations ? 18:38:28 topol, we can do that either as a question in the ops session as one of the starting questions 18:38:31 Hierarchical MultiWine Tasting FTW! 18:38:32 marekd, oui 18:38:35 topol, in fact... thats exactly where I'd put it 18:38:37 stevemar, more line Vin0 18:38:41 how about a Show & Tell for people's strange deployments and extensions 18:38:46 I don't think we've had any sessions on K2K federation 18:38:50 SO I'd vote for a second federation session I think 18:39:05 federate all the things?! 18:39:08 ok so nix VO, and i think split identity is a "meetup topic" 18:39:21 I definitely want to talk about the future policy, but I don't know that we need to use a formal session slot this time 18:39:23 who is bringing the paper 3oz bathroom cups so we can drink alcohol in style??? 18:39:24 I think K2K and the policy discussion will flow into each other anyway 18:39:24 stevemar: lol, federated neutrons, ceilometers, novas etc 18:39:26 nkinder, federation and horizon is another session 18:39:32 integration I mean 18:39:32 stevemar: javak would be in 7th heaven 18:39:40 topol: paper has been nixed IIRC 18:39:43 dolphm, hehe, i 18:39:53 dolphm, i think i'd be scarred of what people would present 18:39:56 stevemar: why would you be interested in VOs ? 18:40:07 morganfainberg: that's why we'll also have wine 18:40:10 VOs also part of K2K 18:40:10 marekd, i'm not, i was just being polite 18:40:15 nkinder dont you know my sarcasm by now? 18:40:17 dolphm, good idea 18:40:20 stevemar, I have a solution...wanna see it? 18:40:29 stevemar: to KENT (?) 18:40:35 http://adam.younglogic.com/2014/10/who-can-sign-for-what/ 18:40:56 so, policy? or more federation? I think we can say "hey people show up on friday and lets talk policy" 18:41:38 ayoung, VO i think is part of K2K and multitenency, probably needs things to be more mature before we can really design around it 18:42:10 morganfainberg, I've been heads down working on just that problem for #moc 18:42:12 morganfainberg, yeah, that's what i was getting at 18:42:28 how do we measure "mature"? just curious 18:42:37 morgainfainberg: if we already have a federation slot, then a policy one seems a good balance 18:42:49 ++ 18:43:08 gyee, considering K2K is expirimental and multitenency isn't merged, i'd say it's a fair bet that neither are mature 18:43:21 :( 18:43:23 sadly, yes 18:43:52 We should have a session on how to get K2K past experimental 18:44:12 bknudson, testing i think is going to be the long and short of it. 18:44:27 to start. 18:44:30 morganfainberg, I'm in the middle of a k2k deployment here 18:44:38 rodrigods, ah nice! 18:44:41 just stopped to maintain HM patches 18:44:50 have several doc improvements suggestions though =) 18:44:58 will submit once I finish it 18:45:00 rodrigods, you will be in paris? 18:45:04 rodrigods: extra 18:45:11 stevemar, no =( 18:45:18 rodrigods, :( 18:45:21 can I participate via hangout from the wine tasting? =D 18:45:27 rodrigods, rock on. rock on. 18:45:36 rodrigods, of course, we will pour you a cup 18:45:40 rodrigods: we will put you on the projector. 18:45:41 rodrigods, that's mandatory 18:45:49 rodrigods, you keep up the good work and we'll mail you a bottle 18:45:50 rodrigods, come on its paris. everyone else is going 18:46:06 =( 18:46:20 haven't got sponsored 18:46:24 Vancouver ftw 18:46:25 =) 18:46:38 I agree that rodrigods is doing a great job :) 18:46:40 15 minutes left 18:46:41 rodrigods, OK, are you astudent? 18:46:46 openstack has scholarship thingy right? 18:46:50 topol, take it out of channel 18:46:52 raildo, you going to be in paris? 18:46:58 gyee read my mind 18:47:00 gyee, topol just finished my masters 18:47:02 not anymore 18:47:04 =) 18:47:04 morganfainberg, yes :) 18:47:09 ok, sorry guys we need to keep on topic 18:47:30 raildo, ok so mind working with gyee to lead the HM session? 18:47:41 morganfainberg, sure 18:48:04 i'll talk with gyee later about this 18:48:14 henrynash, nkinder, ayoung, topol, any of you up to help lead the Ops session (as in help put together the initial questions)? 18:48:21 morganfainberg, I think policy requires some discussion, but probably should be done in the same topic as the token one 18:48:30 sure 18:48:32 morganfainberg: sure, I’d be keen on that one 18:48:37 morganfainberg: yeah, sure 18:48:43 raildo, thanks 18:48:46 ok i'll let you guys come up with some basic questions. 18:48:48 I'm sure we could spend 40 mins talking about tokens and not come to a conclusion 18:48:51 look at what we did last summit 18:49:05 unless somebody's got a great idea that solves all our problems. 18:49:09 morganfainberg, I'm willing to moderate any of the sessions you need me to. 18:49:20 bknudson, we have a session topic for it, dolph has agreed to help lead 18:49:35 gyee, you're welcome 18:49:37 bknudson, but mostly we're focusing on how to work on our current high prio issues 18:49:52 bknudson, and/or examine if there is an alternative to tokens we want to explore. 18:50:14 e.g. signed requests (ec2), etc 18:50:28 ++ 18:50:31 morganfainberg, so probably policy belonds with authorization at [0900 - 0940] Authorization 18:50:53 ayoung, i can shuffle some stuff around and put policy in after that session 18:50:54 morganfainberg: feel free to put me down to help on the hierachical multitenancy too 18:51:01 No, merge it in 18:51:06 henrynash, ++ 18:51:07 ayoung, 40 minutes 18:51:13 ayoung, policy will be mentioned a lot 18:51:15 ayoung, we wont get it all done. 18:51:22 ok. I'm suggesting we're not going to get through token discussion in 40 mins so we can't merge it 18:51:26 henrynash, HM -> revieeeews 18:51:32 bknudson, ++ 18:51:43 rodigods: yeah, I know, I know… 18:51:44 fair enough 18:52:04 so we can either save policy for another time [ let it play out on the ML ] or make it a session. 18:52:15 I'd rather we had a focused discussion and made a decision rather than trying to get through a bunch of topics with no decision on any of them 18:52:17 i'm willing to put the stake in the ground that we'll own that topic as Keystone. 18:52:21 let make it a session 18:52:38 ayoung, ++ 18:52:42 lets put keystone client in the morning, and due auth and policy back to back 18:52:46 ok. we can look at Federation contiuation (non-cross session) as part of the meeting. 18:52:46 morganfainberg: I think it's so tied up with tokens/roles/etc., that we have to own if (or be a large part of it) 18:52:53 nkinder, sure. 18:52:53 I think there's enough discussion to be had on policy topics 18:53:04 for a session by itself 18:53:06 sounds good. 18:53:20 s/meeting/meetup ^ 18:53:39 ok i'll try and setup so client, auth, and policy are grouped together 18:54:04 i'm going to try and get the policy one in the most open of the slots so we can get the most people involved. 18:54:17 what are we doing monday and tuesday if there's no keystone topics? 18:54:31 bknudson: t-shirts hunting 18:54:33 does anyone have any concrete research material we should read up on prior to the summit? 18:54:34 there is cross project stuff on tuesday I think 18:54:40 bknudson, monday is ops track (the big ops sessions, not the project specific one), and tuesday is cross project 18:54:45 Oauth2 has come up a few times, and I'm guessing JWT 18:55:34 bknudson, we could sneak in some informal keystone sessions on monday, as we all walk past each other 18:55:46 ayoung, nkinder, can i have you guys lead the client session if jamie isn't able to? 18:55:50 will there be project pods ? 18:55:54 like in ATL 18:55:59 morganfainberg, any of us can, but Jamie will be there 18:56:01 but we should probably attend other project sessions 18:56:04 marekd, i believe so 18:56:05 marekd, should be 18:56:14 morganfainberg: yeah, I'm sure Jamie will lead it 18:56:20 ayoung, right i just don't want to spring it on him w/o asking 18:56:37 he's expecting to moderate 18:56:40 ok 18:56:41 cool 18:57:14 marekd, ayoung, stevemar, horizon-sso-federation-cli 18:57:21 morganfainberg: i can help 18:57:22 mind if i place you guys on the hook for that one? 18:57:39 I'll take policy, too, if you don't want to morganfainberg 18:58:02 * morganfainberg nods. 18:58:26 hook me for any you need. I'd suspect stevemar and marekd have Federation down, but I've been looking at the Horizon tie in 18:58:37 ayoung, sounds good. 18:58:53 ++ 18:59:04 i'll shuffle the timeslots and see what I can get and hit everyone in -keystone channel before pushing to sched.org 18:59:31 and with that 18:59:33 we're out of time 18:59:40 thanks everyone! 18:59:42 #endmeeting