00:01:00 <thinrichs> #startmeeting CongressTeamMeeting
00:01:01 <openstack> Meeting started Thu Mar 31 00:01:00 2016 UTC and is due to finish in 60 minutes.  The chair is thinrichs. Information about MeetBot at http://wiki.debian.org/MeetBot.
00:01:02 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
00:01:04 <openstack> The meeting name has been set to 'congressteammeeting'
00:01:13 <ramineni_> hi
00:01:19 <masahito> hi
00:01:36 <thinrichs> Hi all
00:02:22 <tsandall> hello
00:02:29 <thinrichs> Agenda is simple: Austin summit and Mitaka release
00:02:33 <thinrichs> Anything else?
00:03:24 <thinrichs> #topic Austin summit
00:03:48 <thinrichs> There's a tentative agenda for our working sessions in Austin…
00:04:37 <thinrichs> We have 4 sessions: 11:00a - 3:20 on Wednesday
00:04:53 <thinrichs> The first is a fishbowl session, where we'll be discussing integrations with other teams.
00:05:22 <thinrichs> The next 3 are working sessions, where we can discuss whatever we need, mainly within our own team.
00:06:02 <thinrichs> We'll need to decide what we want to discuss in those 3 sessions.
00:06:06 <thinrichs> Any suggestions?
00:07:24 <thinrichs> Here are some thoughts…
00:08:07 <thinrichs> (i) distributed architecture, (ii) HA architecture built on top of distributed arch
00:08:56 <thinrichs> We could also have a session where we do some more long-range planning
00:09:13 <ekcs> sounds good.
00:09:15 <thinrichs> A bug-squashing session.
00:10:34 <thinrichs> We should all think about what would be most beneficial to discuss over the next couple weeks.
00:10:59 <thinrichs> Ideal topics are those where we've all had some time to work/think about them before the summit, so the middle of a conversation, not the beginning.
00:11:14 <thinrichs> So let's all take an action item to give that some thought over the next couple of weeks.
00:11:29 <thinrichs> One other thing that is related to the fishbowl session.
00:11:45 <thinrichs> I've been helping a group of people who want a policy-enabled Nova scheduler.
00:12:06 <thinrichs> We put together a spec and plan to discuss that spec as part of the fishbowl session.
00:12:12 <thinrichs> #link https://review.openstack.org/#/c/297898/
00:12:54 <thinrichs> The idea right now is to have the scheduler use Congress's datalog engine, which we'd need to pull out into a library.
00:13:50 <thinrichs> Admins write policies that describe which hosts a new VM request is permitted to be scheduled on.
00:14:01 <thinrichs> The policies would be written either in Datalog or a simple fragment of Datalog codified in YAML.
00:15:29 <thinrichs> This effort is related to the delegation functionality we've worked on a bit in the past.
00:16:12 <thinrichs> There are still details to work out about where the policy lives, how it gets updated, and so on.
00:16:34 <thinrichs> Any questions or comments?
00:17:19 <ekcs> sounds great.
00:17:24 <masahito> sounds nice.
00:18:00 <ekcs> I’ve been wanting to make a YAML embedding of Congress datalog. and letting congress accept policy in YAML.
00:18:25 <ekcs> Maybe I’ll be able to borrow/help from this effort.
00:18:45 <thinrichs> ekcs: In that spec we only have  a *fragment* of Datalog encodable in YAML.  I tried a full encoding, but no one liked it b/c it was still datalog.
00:18:57 <thinrichs> ekcs: It'd be great to have help.
00:20:02 <ekcs> haha i see.
00:20:06 <thinrichs> The spec asks for feedback for a couple of things, so I'm planning to hold off on doing any implementation until we understand what the Nova community thinks and what the parameters are.
00:22:20 <thinrichs> I'm excited at the prospects, but I'm also waiting to see what the Nova scheduling team thinks.
00:22:25 <thinrichs> Anything else for Austin?
00:23:29 <thinrichs> #topic Mitaka
00:23:55 <thinrichs> A few important dates (from the release team):
00:23:57 <thinrichs> #link http://lists.openstack.org/pipermail/openstack-dev/2016-March/090877.html
00:24:24 <thinrichs> 1. Tomorrow is the final day for a release candidate
00:24:49 <thinrichs> 2. Friday (2 days from now) is the last day to request a full release
00:25:00 <thinrichs> 3. Early next Thursday, final release happens
00:25:22 <thinrichs> What this means is that we need to merge any outstanding fixes by tomorrow.
00:25:43 <thinrichs> Any fixes after that will only be allowed in if they are Critical.  Release team has final say.
00:26:37 <ekcs> Got it.
00:26:40 <thinrichs> For us, we have 1 decision to make about the release
00:26:42 <ekcs> I'd like to figure out what to do with the keystone authentication issue. I think it's important to get some fix in for Mitaka.
00:26:53 <thinrichs> Yep.
00:27:25 <thinrichs> ekcs, ramineni: does one of you want to lay out the decision we need to make?
00:27:40 <ekcs> sure.
00:28:37 <ramineni_> thinrichs: i thought we could go for sessions only for problematic clients and deal with others later in N release
00:28:58 <thinrichs> ramineni_: That sounds good to me.
00:28:59 <ramineni_> thinrichs: but just going through ekcs comments on the patch
00:29:40 <ramineni_> thinrichs: i think there shouldn't be any problems in using sessions if client supports it ..
00:30:12 <ekcs> For some of the datasource drivers, a workaround is needed where Congress manually refreshes the auth token after it expires. But the workaround is broken for some drivers (depends on the corresponding client)
00:30:13 <thinrichs> Right now we think the datasources we need to fix are Glance, Heat, Murano.  Correct?
00:31:07 <ramineni_> thinrichs: glance heat , im able to reproduce in my env .. murano havent checked yet
00:31:25 <ekcs> I have fixes that instantiate a new client each time the token expires. https://review.openstack.org/298960
00:31:26 <ekcs> ramineni_ has a fix that uses a new a better model (keystone session) for authentication. https://review.openstack.org/299121
00:31:46 <ekcs> we mostly agree that in the future the session model should be used. question is what to do for mitaka.
00:31:58 <ekcs> thinrichs: yes those three.
00:33:22 <ramineni_> ekcs: mitaka also, how about using sessions only? there shouldnt be any problems right , as we are usig their functionality directly
00:34:11 <thinrichs> ekcs did a good thing in preparing both versions so we can make the decision and merge today, if we need to.
00:34:17 <thinrichs> So we just need to make a decision.
00:34:38 <thinrichs> I think we agree that the keystone sessions are the right long-term play.
00:34:51 <thinrichs> What's the possible downside to using sessions in Mitaka?
00:35:25 <thinrichs> Bugs in other peoples' code that we don't know about b/c we haven't tested thoroughly.
00:35:28 <thinrichs> Anything else?
00:35:29 <ekcs> One issue with sessions, I couldn’t get it to work for murano. But if we can fix that right away then that’s an option for mitaka. https://review.openstack.org/299683
00:36:12 <thinrichs> Are we thinking that for Glance/Heat, we should definitely go with sessions then?
00:37:40 <ekcs> thinrichs: that and the potential that we’re not quite using it correctly. I haven’t been able to find detailed docs for the clients, so I’m just doing trial and error at this point swithing to sessions. With the client instantiation workaround, it’s all the same code we’ve been using so less risk of introducing a problem because we’re not using the client the way they expect.
00:38:19 <ekcs> thinrichs: I’m in favor of all one or the other. should make future mitaka support easier.
00:40:33 <thinrichs> I'm thinking that since the Glance/Heat seems to be working with sessions, we should merge it.
00:41:00 <ekcs> thinrichs: I see.
00:41:12 <thinrichs> If we can get the Murano one working by tomorrow morning in the US, we merge it too.
00:41:35 <thinrichs> Otherwise, we use the other Murano fix.
00:42:08 <thinrichs> ekcs: Would you prefer to do all in keystone sessions or none?
00:42:24 <ekcs> ramineni_: how did you figure out exactly what to pass to client instantiation? seems heat takes endpoint but glance does not. I’m trying to get a sense of how confident we are about how we’re calling the client using session.
00:43:00 <ekcs> thinrichs: If it’s just murano I’m okay with splitting.
00:43:35 <ramineni_> ekcs: i have checked the client code , what parameters it accepts
00:44:33 <ekcs> got it. ramineni_ if you feel confident about it then I’m okay with using session for glance and heat.
00:45:01 <ekcs> it worries me a little bit why glance doesn’t take endpoint. like what if by not providing endpoint something goes wrong later.
00:45:30 <thinrichs> ekcs: when you tried the sessions with Glance, did you see the same problem you saw before?
00:45:37 <thinrichs> No, right?
00:46:00 <ekcs> Right. my testing says it works.
00:46:09 <ramineni_> ekcs: using keystone sessions is straight forward, I have used the same example code in keystoneclient itself
00:46:31 <thinrichs> Could always be another, different issue down the road that we just haven't waited long enough to see.
00:47:16 <thinrichs> But I think that with that kind of bug, we probably wouldn't catch it even if we merged this patch a month ago.
00:47:23 <ekcs> I’m happy with using session. It just scares me when we’re deducing from code how to call client instantiation. but well.
00:48:06 <thinrichs> ekcs: makes sense.  It's good that you raise it.
00:48:42 <ekcs> I’m almost exactly on the fence. So I think we should do sessions since thinrichs and ramineni_ like that.
00:49:09 <ekcs> I would’ve argued muchmore strongly for staying with the workaround if the action execution wasn’t an issue.
00:49:29 <thinrichs> ekcs: action-execution is an issue?  I missed that somehow?
00:49:42 <ramineni_> ekcs: me too..action-execution?
00:49:58 <ekcs> https://bugs.launchpad.net/bugs/1564152
00:50:00 <openstack> Launchpad bug 1564152 in congress "When 1 driver fails to load, all drivers fail to list (Error 500)" [Undecided,New]
00:50:08 <ekcs> oh wrong bug.
00:51:07 <ekcs> https://bugs.launchpad.net/congress/+bug/1564115
00:51:08 <openstack> Launchpad bug 1564115 in congress "auth_token expiry on action execution (Glance, Murano)" [Undecided,New]
00:51:58 <ekcs> session would fix this issue. extending the client recreate workaround to this issue would be ugly.
00:52:04 <thinrichs> Agreed
00:52:27 <thinrichs> I was actually barely on the side of the sessions before.  This pushes me more toward sessions.
00:52:46 <ramineni_> ekcs: did u test with sessions
00:53:02 <ekcs> I guess we have a decision then.
00:53:19 <ekcs> ramineni_: no I didn’t test with sessions. but sessions should fix it.
00:53:26 <ramineni_> ekcs: action execution, i didnt do any testing related to that ..so
00:54:00 <ekcs> right that’s a worry. but authentication *should* be the same.
00:54:15 <thinrichs> Sounds like we want to do 2 things before tomorrow morning...
00:54:30 <thinrichs> test that action-execution works for Glance/Heat with sessions
00:54:52 <thinrichs> and fix up Murano to either use sessions properly.
00:55:30 <thinrichs> or to use the workaround (which ekcs finished).
00:55:31 <ramineni_> thinrichs: ok
00:55:38 <thinrichs> Sorry…that was unclear:
00:55:47 <ekcs> ok
00:55:48 <thinrichs> (i) Test action-execution for Glance/Heat with sessions.
00:56:15 <thinrichs> (ii) Fix up Murano to use sessions or abandon that and use the workaround. And then test polling and action-execution.
00:56:47 <thinrichs> ramineni_, ekcs: do either of you have time for those?
00:57:22 <ramineni_> thinrichs: ill check (ii) , ekcs, can u do quick testing on i, as you have env?
00:57:40 <ekcs> I can work on testing. but I’m not sure I’ll be able to do action execution test on heat.
00:57:59 <thinrichs> ekcs: why the problem with Heat?
00:58:01 <ekcs> I’ll give it a go.
00:58:45 <ekcs> The only action I could find was start a new stack. and I haven’t been able to get that working.
00:58:51 <thinrichs> ramineni_: if you finish up with Murano soon, maybe ekcs will still have time to test it.
00:58:56 <thinrichs> ekcs: makes sense
00:59:12 <ekcs> but at least I should be able to check that the client attempted the action execution and didn’t run into auth issue.
00:59:17 <ekcs> even if the action failed.
00:59:27 <thinrichs> 1 minute remaining.  Let's hop onto #congress to finish up quickly.
00:59:29 <ramineni_> thinrichs: sure, i will check on it, as soon as i reach office
00:59:49 <ramineni_> thinrichs: dont have test env now
01:00:00 <thinrichs> ramineni_: Great!  It's worth a shot at having ekcs do the testing at least.  Otherwise, we may not have time.
01:00:14 <thinrichs> ramineni_: How long until you think you'll have something?
01:00:38 <thinrichs> ekcs: If we can't test Heat action execution, that's fine.
01:00:57 <ekcs> ok see you on #congress.
01:01:00 <thinrichs> ekcs: maybe if you have some spare time, and ramineni_ hasn't finished the Murano session, you could test the Murano workaround.
01:01:02 <thinrichs> Out of time.
01:01:04 <thinrichs> Thanks all!
01:01:07 <thinrichs> #endmeeting