15:00:36 <n0ano> #startmeeting gantt
15:00:37 <openstack> Meeting started Tue Dec 16 15:00:36 2014 UTC and is due to finish in 60 minutes.  The chair is n0ano. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:38 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:41 <openstack> The meeting name has been set to 'gantt'
15:00:51 <n0ano> anyone here to talk about the scheduler?
15:00:55 <alex_xu> \o
15:00:55 * bauzas waves
15:00:58 <bauzas> \o
15:01:08 <edleafe> o/
15:01:12 <n0ano> bauzas, you always beat me by about 10 seconds :-)
15:01:37 <bauzas> n0ano: eh, my middle name is Tag Heuer
15:02:06 <n0ano> bauzas, mine is `whenever` :-)
15:02:12 <n0ano> anyway, let's start
15:02:20 <n0ano> #topic code clean up status
15:02:23 <PaulMurray> hi
15:02:42 <n0ano> I saw you had a long IRC converstation with jaypipes , did you work everything out?
15:02:53 <bauzas> by you, whose you're talking ? :)
15:03:05 * bauzas likes long convos with jaypipes
15:03:21 <n0ano> you == bauzas , I think you were talking about the isolate scheduler DB
15:03:25 <bauzas> s/whose/whom (darn French !)
15:03:39 <bauzas> n0ano: oh
15:03:53 <n0ano> bauzas, sorry, s/whom/who (I can be pedantic about grammar too)
15:04:00 <bauzas> n0ano: well, everything is normal, he doesn't like my proposal, I consequently hassled him
15:04:18 <bauzas> n0ano: nah, enough kidding
15:04:44 <bauzas> n0ano: so, my former proposal was about to persist aggregates information in a separate table held by the Scheduler
15:05:11 <bauzas> n0ano: as I had many reviews saying it was a bad idea, I consequently changed my spec to only provide an in-memory update
15:05:37 <bauzas> n0ano: there are still some concerns about how to populate that cache, so I still expect further feedback
15:05:38 * edleafe is glad he didn't update his spec for table persistence
15:05:47 <n0ano> ahh, that was the memory cache you two were talking about, is that in the updated spec I just saw the notic for?
15:05:57 <bauzas> edleafe: hence my comment to wait the aggregate spec to be at least one +2'd ;)
15:06:09 <edleafe> bauzas: exactly :)
15:06:10 <bauzas> n0ano: correct
15:06:20 <bauzas> with the link, it's better
15:06:22 <bauzas> #link https://review.openstack.org/#/c/89893/
15:06:54 <bauzas> above is the updated spec taking in account all previous comments
15:07:19 <n0ano> so, definitely reviewable but sounds like you expect at least one more update on it
15:07:25 <bauzas> that's not perfect IMHO, as the Scheduler is still calling the Aggregate API once in its lifetime, but that's a good path till we split the scheduler
15:07:43 <bauzas> n0ano: well, you know that the devil is in the details...
15:08:35 <bauzas> anyway, I'm preparing to request an exception for the freeze if we can't make it for Thursday
15:08:50 <bauzas> that's a very old spec and still necessary for the split
15:09:01 <n0ano> bauzas, good plan, i'm guessing we'll need it but we should be close
15:09:19 <bauzas> n0ano: yeah, the main problem is to define how to do this
15:09:43 <bauzas> n0ano: once we agree on the plan for updating the Scheduler view and all the tradeoffs, then we can quickly iterate
15:09:53 <alex_xu> bauzas, how can remove the aggregate api calling when we split scheduler?
15:10:21 <alex_xu> pull data instead of push data?
15:10:23 <bauzas> alex_xu: well, that's something which can be discussed
15:10:43 <bauzas> alex_xu: I wrote in the alternatives sections that it's possible to persist the dataz
15:10:59 <alex_xu> bauzas, ok, I will read it
15:11:03 <bauzas> alex_xu: but it would require a separate nova scheduler service
15:11:35 <alex_xu> bauzas, looks like our api is push data, this won't changed
15:11:37 <bauzas> anyway, that actually comes to the fact that we plan to ship first a library
15:12:00 <bauzas> if we consider a library, that doesn't work well with the concept of having its own DB
15:12:28 <bauzas> anyway, I think it's a bit early for discussing this - at least until this 89893 spec merges
15:12:54 <alex_xu> ok, think about it later
15:13:01 <bauzas> because that spec is necessary for thinking about how we can ship the scheduler
15:13:04 <n0ano> bauzas, I agree, get the overall design approved and then work out the details
15:13:42 <bauzas> n0ano: so, that's it for this specific BP
15:13:53 <n0ano> cool, tnx
15:13:56 <bauzas> n0ano: that's good we targeted it for K3 anywauy
15:14:14 <bauzas> because it seems it will probably a couple of weeks still to get it merged
15:14:29 <n0ano> looking at our dashboard, https://wiki.openstack.org/wiki/Gantt/kilo#Tasks
15:15:05 <n0ano> the isolate scheduler DB is open but we have line of sight for approval,
15:15:21 <n0ano> jaypipes, his object model spec is the big open
15:15:24 <bauzas> agreed
15:15:30 <bauzas> +1 too
15:15:42 <bauzas> and I have some concerns about what he's planning to do
15:16:04 <n0ano> unfortunately, he doesn't seem to be responding today
15:16:09 <bauzas> I actually required many review iterations to understand what the spec was doing, so that's my bad if I'm giving -1 so late
15:16:37 <n0ano> bauzas, no worries, that's just part of the process
15:17:12 <bauzas> I'm not a logical person, that means I need sometimes more time to understand some concepts :)
15:17:42 <n0ano> I think I'll try and find jaypipes on IRC later and see if we can get an update on his plans for that spec
15:18:32 <n0ano> of course, I'm kind of avoiding the issue that now we have to implement the specs that have been approved
15:18:40 <bauzas> PaulMurray: did you get updates from jaypipes about his spec ?
15:19:08 <bauzas> PaulMurray: I saw you were discussing about it, even by the big G hangout stuff
15:19:56 <bauzas> sounds like we also lost PaulMurray
15:20:16 <n0ano> yep (I've had network issues in the past myself)
15:20:23 <bauzas> netsplit ?
15:20:37 <n0ano> we'll just have to ping jaypipes later
15:20:48 <bauzas> agreed
15:21:17 <n0ano> we've kind of covered the specs let's move on
15:21:22 <n0ano> #topic opens
15:21:37 <n0ano> anything else we want to discuss today?
15:21:56 <bauzas> nothing really for me
15:22:02 <edleafe> nor me
15:22:13 <n0ano> short meeting works for me
15:22:17 <bauzas> \o/
15:22:19 <alex_xu> nothing from me
15:22:35 <n0ano> we still have reviews so let's all do that
15:22:53 <edleafe> sounds like a plan
15:22:56 <n0ano> BTW, I'll be around next week (big holiday coming up I hear) do we still want a meeting next week?
15:23:05 <bauzas> n0ano: oh good point
15:23:16 <edleafe> I'll be around
15:23:16 <bauzas> lemme check the date
15:23:25 <edleafe> but yeah, a lot of people will be off
15:23:27 <bauzas> sounds feasible
15:23:35 <n0ano> Christmas day is Thurs so Tues is fine by me
15:23:55 * n0ano has all his shopping done for the first time ever
15:24:01 <bauzas> I should be off by end of Tues till 5th Jan
15:24:31 <n0ano> bauzas, NP, enjoy, I'll still run a meeting next week, it could be `very` short
15:24:37 <bauzas> n0ano: ack
15:24:47 <PaulMurray> bauzas, I was pulled into something by my manager - I can update at the end if you like
15:24:56 <n0ano> PaulMurray, go for it
15:25:22 <PaulMurray> bauzas, ok
15:25:52 <PaulMurray> I have a discussion with jaypipes about his spec for resource object models
15:26:04 <bauzas> right
15:26:15 <PaulMurray> The basic point was making sure that two things are covered that were requirements for ERT
15:26:29 <PaulMurray> 1) ability to select resources for different virt drivers
15:26:47 <PaulMurray> 2) making sure it is easy to develop and contribute new resource objects
15:27:07 <PaulMurray> I also noted a bug in the resource tracker to do with ironic
15:27:18 <n0ano> PaulMurray, good goals, I approve
15:27:31 <PaulMurray> Ironic uses resources differently to other virt drivers
15:27:45 <PaulMurray> and that is not accounted for at the moment
15:27:54 <bauzas> PaulMurray: I agree, I had many troubles with Ironic when implement detach-service
15:28:12 <PaulMurray> so this is an opportunity to make sure those kind of things are easy to deal with in the future
15:28:30 <PaulMurray> I can find the link for the bug - hold on
15:28:59 <PaulMurray> https://bugs.launchpad.net/nova/+bug/1402658
15:29:01 <uvirtbot> Launchpad bug 1402658 in nova "resource tracking is incorrect for ironic" [Undecided,New]
15:29:43 <PaulMurray> That can be dealt with as a bug fix, but dealing with resources that behave differently should be inherent in the solution
15:29:59 <PaulMurray> Anyway - Jay is going to do a revision to day I think
15:30:27 <bauzas> PaulMurray: I was also having some concerns about how he's planning to update the stats
15:31:02 <bauzas> PaulMurray: IIUC, he was planning to update the UsageSpec classes, but not the ComputeNode object
15:31:59 <PaulMurray> bauzas, I got a little lost on that part - I need to get more familiar with the proposed code
15:32:26 <n0ano> bauzas, as long as you indicate that in your review that should make jaypipes address those issues
15:32:39 <bauzas> PaulMurray: I think that's the main problem I had
15:32:50 <bauzas> PaulMurray: I was a little lost of what was the actual plan
15:33:17 <bauzas> but let's jaypipes clarify that in his spec
15:33:18 <PaulMurray> bauzas, I know that some of it looks confused because it deals with legacy around the ComputeNode object
15:34:00 <bauzas> PaulMurray: my main problem is that there are many set of classes and I was still having some problems with seeing the relationship with the ComputeNode object
15:34:11 <PaulMurray> bauzas, I'm not sure its done cleanly enough yet, but the path from legacy to being clean is not quite there for me
15:34:21 <bauzas> PaulMurray: agreed
15:34:24 <PaulMurray> hence my concerns about ease of use
15:34:38 <PaulMurray> or rather ease of extension
15:34:45 <bauzas> PaulMurray: yeah, I just want to make sure it's quite understandable and readable by most of us
15:34:50 <n0ano> I'm concerned about timing, even with an extension we don't have much time to get this right
15:35:28 <PaulMurray> n0ano, yes, I said would make myself available to help get it right
15:35:44 <n0ano> PaulMurray, tnx, anything to help
15:35:46 <PaulMurray> n0ano, it might need a restricted scope and a statement about where it intends to go
15:36:11 <PaulMurray> n0ano, can't really say much in jaypipes absense
15:36:29 <n0ano> yeah, understood, we'll try and get him later on IRC
15:36:35 <bauzas> agreed
15:36:52 <n0ano> so, unless there's anything else
15:37:33 <n0ano> I'll thank everyone and we'll talk again next week
15:37:40 <bauzas> sure thing
15:37:42 <n0ano> #endmeeting