14:59:37 <n0ano> #startmeeting gantt
14:59:38 <openstack> Meeting started Tue Apr  1 14:59:37 2014 UTC and is due to finish in 60 minutes.  The chair is n0ano. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:59:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:59:42 <openstack> The meeting name has been set to 'gantt'
14:59:53 <n0ano> anyone here want to talk about the scheduler?
15:00:17 <bauzas> o/
15:01:04 <bauzas> btw. please note France switched to CEST yesterday
15:01:08 <lcostantino> o/
15:01:10 <mspreitz> hi
15:01:29 <n0ano> bauzas, you have my sympathy, I `hate` daylight savings time :-)
15:01:29 <bauzas> so, there could be some possibilities that people miss the meeting
15:01:47 * johnthetubaguy raises late hand, that is slightly absent
15:01:54 <bauzas> well, that's now 5pm now in France, I prefer CET :D
15:02:08 <n0ano> hence OpenStack meetings are always UTC, the meeting follow the sun :-)
15:02:27 <n0ano> anyway, to begin
15:02:33 <n0ano> #topic no db scheduler
15:02:36 <bauzas> sure
15:03:02 <n0ano> given that this project is effectively deferred until after Juno it's silly to try and talk about it every week...
15:03:19 <bauzas> just a fyi, a recent concern about querying host aggregates has been raised where we identified the need of storing AZs and aggs within ComputeNode state
15:03:38 <bauzas> so, no-db scheduler could possibly be impacted by this
15:03:51 <n0ano> at the same time I don't want to forget about it so I think we'll put a note about it as a postscript to the agenda but don't need to talk about it
15:04:03 <johnthetubaguy> well no-db is blocked by the scheduler split at this point, I feel
15:04:19 <johnthetubaguy> its just the AZ info would go into mem-cache, so not sure thats a big one
15:04:19 <n0ano> bauzas, so, this is just another case of it's going to wait until after Juno
15:04:32 <johnthetubaguy> n0ano: I don't think it has too
15:04:35 <bauzas> johnthetubaguy: indeed, just a fyi again :)
15:04:45 <johnthetubaguy> n0ano: only till after that lib is created
15:04:52 <PaulMurray> Hi all, I seem to have come in half way through a conversation
15:04:57 <johnthetubaguy> n0ano: then they can use that seam to make it optional
15:05:11 <n0ano> johnthetubaguy, I was referring to no-db, not the aggregates issue
15:05:33 <johnthetubaguy> n0ano: yep, I was meaning thats the solution for no-db, wait for the seam, then use it, so its optional
15:05:37 <bauzas> anyway, boris-42 told us there were some resource issues for working on no-db, so it will have to wait until Juno :)
15:05:50 <johnthetubaguy> bauzas: it is Juno
15:05:56 <boris-42> bauzas hi
15:06:02 <bauzas> hi boris-42
15:06:03 <n0ano> I think we're in violent agreement
15:06:06 <boris-42> bauzas yep it will wait for juno=)
15:06:09 <boris-42> bauzas lol=)
15:06:25 <n0ano> boris-42, still think it's important so we don't want to forget about it
15:06:31 <johnthetubaguy> sure, juno is open now
15:06:44 <bauzas> johnthetubaguy: well, ok
15:06:48 <boris-42> n0ano hehe=)
15:06:52 <johnthetubaguy> needs nova-specs sorting, etc
15:07:06 <boris-42> n0ano sorry we have some troubles with Alexei
15:07:18 <boris-42> n0ano I think I will probably finish that work
15:07:26 <boris-42> n0ano to avoid unexpected situation this time..
15:07:35 <boris-42> n0ano was too busy to work on it during icehouse =(
15:07:58 <n0ano> boris-42, NP, we'll just keep pushing you as the owner, who you get to do the work is your problem
15:08:12 <bauzas> boris-42: we just discussed about it and said it was worth waiting for sched-lib to be delivered before committing on no-db sched
15:08:13 <boris-42> n0ano lol=)
15:08:29 <boris-42> bauzas yep it is hard to make lib without it..
15:08:36 <boris-42> bauzas cause you are totally bind to nova data
15:08:50 <boris-42> bauzas and structuree of nova data
15:08:56 <bauzas> boris-42: we do baby-steps :)
15:09:03 <boris-42> bauzas hehe=)
15:09:06 <boris-42> bauzas yep yep
15:09:06 <bauzas> boris-42: at the moment, we work on the interfaces
15:09:21 <bauzas> boris-42: but we need to chase all calls from within the scheduler indeed
15:09:24 <boris-42> bauzas so you would like to use RPC as interface?
15:09:56 <boris-42> bauzas I think that interface e.g. (hostname, namespace, values) is the best one
15:10:01 <bauzas> boris-42: well, I'm teasing the 2nd topic, but we're committed to deliver a sched client
15:10:27 <bauzas> boris-42: which will take use of the sched rpcapi
15:10:45 <bauzas> boris-42: baby step, proxying :)
15:10:46 <boris-42> bauzas I think that it's more important what you have inside
15:10:51 <bauzas> boris-42: indeed
15:11:02 <bauzas> boris-42: but we just need to perform one-by-one :)
15:11:07 <johnthetubaguy> the key point I was thinking, we can make no-db optional right?
15:11:09 <boris-42> bauzas sure sure
15:11:24 <boris-42> johnthetubaguy it changes.. the way how scheuler works
15:11:40 <n0ano> johnthetubaguy, it was my understanding that was the goal
15:11:42 <boris-42> johnthetubaguy so there is no magic button "turn it off"
15:11:47 <johnthetubaguy> boris-42: I still think it should be an option, new no-db stuff, or current db stuff
15:12:04 <boris-42> johnthetubaguy one more separated project?)
15:12:15 <bauzas> boris-42: nah, just a conf flag :)
15:12:22 <boris-42> bauzas it has different arch
15:12:24 <n0ano> bauzas, +1
15:12:28 <johnthetubaguy> boris-42: well we are creating gantt, but yeah, just a config flag for this case
15:12:42 <boris-42> johnthetubaguy it will be different code with different logic..
15:12:50 <boris-42> johnthetubaguy so only interface could be the same
15:12:59 <johnthetubaguy> boris-42: best to take it through the nova-specs design review anyways, before the blueprint will be approved
15:13:03 <boris-42> johnthetubaguy but in that case we should discuss what "interefact" we would like
15:13:14 <boris-42> interface**
15:13:15 <johnthetubaguy> boris-42: right, thats the scheduler lib
15:13:17 <bauzas> johnthetubaguy: +1, god loves nova-specs :D
15:13:33 <boris-42> johnthetubaguy so what I need in no-db is to get updates from host in next way
15:13:48 <bauzas> boris-42: sounds like we should do peer reviews
15:13:50 <boris-42> (hostname, namespace, values)
15:13:57 <johnthetubaguy> the idea is to call some python code with that above interface, right
15:14:10 <johnthetubaguy> then the old stuff writes to db
15:14:12 <bauzas> boris-42: because I will proxy the call from the resource tracker
15:14:12 <n0ano> bauzas, I'm thinking we need to hash this out at Atlanta
15:14:23 <bauzas> n0ano: strong +2 here
15:14:25 <johnthetubaguy> your new scheduler driver thingy, does what it needs to do (RPC or whatever)
15:14:59 <boris-42> johnthetubaguy the problem is one more we are chaining how it works
15:15:05 <boris-42> johnthetubaguy it's not only driver part
15:15:07 <bauzas> again, we're teasing the summit topic :)
15:15:13 <boris-42> johnthetubaguy it's everything
15:15:23 <boris-42> johnthetubaguy storing results, filters, processing results, api
15:15:31 <n0ano> bauzas, getting issues out early is good, resolving might have to wait for the summit
15:15:47 <boris-42> johnthetubaguy there is only one common possible place
15:16:05 <bauzas> n0ano: my bad, I was saying we're teasing today's meeting topic called 'summit talks" :D
15:16:36 <johnthetubaguy> boris-42: I think I get the bits you are talking about, but it really needs to be an option, I feel, so people can choose when to migrate to the new method
15:16:46 <n0ano> bauzas, +1 :-)  (I don't mind non-linear converations :-)
15:17:00 <johnthetubaguy> its about getting the correct seams in the current system, I think
15:17:02 <boris-42> johnthetubaguy heh let me think about it
15:17:13 <johnthetubaguy> OK
15:17:15 <boris-42> johnthetubaguy how to make it optional=)
15:17:25 <boris-42> johnthetubaguy separated project seems like a good idea lol=)
15:17:26 <johnthetubaguy> I think its about a replacement hostmanager, sort of
15:17:34 <boris-42> johnthetubaguy yep whole stuff..
15:17:36 <johnthetubaguy> for no-db at least
15:17:39 <n0ano> boris-42, I think optional is important, even if it means changing your design a little
15:18:06 <boris-42> n0ano the issue that it changes everything=)
15:18:08 <bauzas> +1
15:18:25 <bauzas> boris-42: I don't remember seeing such a big bang :)
15:18:26 <boris-42> n0ano so if you would like to make it optimal in gantt we should have common API
15:18:34 <mspreitz> Does it even change nova clients?
15:18:39 <boris-42> mspreitz lol
15:18:47 <boris-42> mspreitz nope
15:19:02 <bauzas> it's worth reading the spec here :)
15:19:04 <boris-42> mspreitz but it changes nova.compuetes
15:19:06 <mspreitz> Heat wants to get out of the business of maintaining its own cache of state...
15:19:34 <boris-42> bauzas btw I made session
15:19:46 <boris-42> bauzas so I can describe whole architecture
15:19:47 <bauzas> boris-42: and I replied :)
15:19:59 <boris-42> bauzas and seems like I should find time to update docs and learn what was done in Gantt
15:20:47 <n0ano> boris-42, nothing new has been done in gantt (yet), it is just a forklift of the current scheduler design
15:20:53 <bauzas> boris-42: ok, let's discuss this in the second next topic :)
15:21:04 <boris-42> bauzas heh=)
15:21:11 <boris-42> bauzas n0ano ok I have to go now=)
15:21:20 <n0ano> good seque
15:21:24 <boris-42> bauzas n0ano  but I will try to organize development of this stuff
15:21:31 <boris-42> from Mirnatis side
15:21:33 <n0ano> #topic scheduler forklift
15:21:42 <bauzas> boris-42: just take a look at my comment to your proposal in the summit page :)
15:21:50 <bauzas> boris-42: and ping me later if needed :)
15:22:06 <boris-42> bauzas ok thanks
15:22:24 <bauzas> ok, so about sched forklift
15:22:27 <bauzas> a few pointers
15:22:37 <bauzas> https://review.openstack.org/#/c/82133/
15:22:48 <bauzas> that's the nova-specs bp
15:22:55 <bauzas> feel free to take a look on it
15:23:05 <johnthetubaguy> +1 please review
15:23:29 <bauzas> also
15:23:38 <bauzas> draft implementation is in progress
15:23:43 <bauzas> https://review.openstack.org/#/c/82778/
15:23:51 <bauzas> still need to adapt based on BP review of course
15:24:05 <bauzas> and still chasing up issues with unittests :)
15:24:28 <bauzas> woah, I was playing with climate CI, this one is huuuge :)
15:24:52 <bauzas> I really enjoy having 200+ tests failing :)
15:25:18 <n0ano> bauzas, that doesn't bother me, it normally means something simple and basic is wrong
15:25:18 <bauzas> hence the draft, so as to ramp-up on nova
15:25:46 <n0ano> bauzas, it's the 3 or 4 random failures that cuase headaches
15:25:55 <bauzas> n0ano: :)
15:26:26 <n0ano> note, I have a tendency to not review things until the tests pass, you never know what might need to change
15:26:41 <bauzas> n0ano: then focus on the BP :)
15:26:51 <bauzas> n0ano: Jenkins gave me +1 :)
15:27:05 <n0ano> on a BP, no bonus points for that :-)
15:27:45 <bauzas> well, nothing to add here :)
15:28:00 <bauzas> waiting reviews on the BP and in progress on the draft implem
15:28:02 <n0ano> sounds like we all need to review and then we can discuss more intelligently
15:28:14 <bauzas> n0ano: sounds a good option to me :D
15:28:16 <n0ano> bauzas, cool, tnx for the work!
15:28:29 <n0ano> moving on then
15:28:36 <n0ano> #Atlanta sessions
15:28:39 <bauzas> nah
15:28:42 <bauzas> #topic :)
15:28:52 <n0ano> #topic Atlanta sessions
15:28:56 * n0ano sign
15:28:56 <bauzas> :)
15:29:01 <n0ano> s/sign/sigh
15:29:04 <n0ano> anyway...
15:29:16 <n0ano> I search and found 4 proposals so far...
15:29:30 <bauzas> n0ano: +1, I found 4
15:29:41 <n0ano> No-db scheduler (boris put the same proposal in twice, I guess he really wants it :-)
15:29:45 <bauzas> n0ano: 2 of me, 1 from boris-42, one from jay lau
15:29:56 <n0ano> sched hints for VM life cycle (looks good to me)
15:30:06 <n0ano> Forklift & gantt APIs
15:30:23 <n0ano> Does anyone else have plans they know of for more proposals?
15:30:39 <bauzas> we should hurry up for cross-project sessions :)
15:30:40 <mspreitz> BTW, what is the URL for proposals?
15:30:50 <bauzas> mspreitz: lemme give them to you
15:30:59 <bauzas> http://summit.openstack.org/cfp/details/180  : no-db
15:31:14 <bauzas> http://summit.openstack.org/cfp/details/99 : sched hints
15:31:23 <n0ano> mspreitz, for submitting them...
15:31:26 <bauzas> http://summit.openstack.org/cfp/details/80 : sched forklift
15:31:28 <n0ano> #link http://summit.openstack.org/
15:31:41 <bauzas> http://summit.openstack.org/cfp/details/140 : gantt API
15:31:55 <bauzas> there was a thread in -dev
15:32:21 <bauzas> stating that proposals for cross-projects should be sent by end of ...
15:32:24 <mspreitz> page title says "icehouse"
15:32:38 <n0ano> threads in -dev can easily get lost in the noise
15:32:42 <n0ano> mspreitz, you sure, its
15:32:43 <bauzas> mspreitz: that's an April joke :)
15:32:52 <n0ano> mspreitz, you sure, it says Juno in my browser
15:33:00 <bauzas> n0ano: +1
15:33:09 <mspreitz> ah, it's my browser, inheriting something from before
15:33:24 <n0ano> mspreitz, reload is your friend (I get hit with that all the time)
15:33:28 <bauzas> got the email
15:33:34 <bauzas> lemme give you the link
15:33:49 <mspreitz> nvm
15:33:56 <mspreitz> It's Firefox's bookmarking code
15:34:05 <mspreitz> don't waste meeting time on this
15:34:25 <n0ano> so, the 4 we have outstanding look good to me but if there are others people want to propose they should do it soon
15:34:51 <bauzas> #link http://lists.openstack.org/pipermail/openstack-dev/2014-March/031384.html
15:35:15 <bauzas> ETA April 10th
15:35:18 <mspreitz> the listing page has way more than 4 for nova
15:35:40 <n0ano> mspreitz, well yeah, I'm only concerned about scheduler/gantt
15:35:49 <bauzas> for submitting cross-project and "other projects" proposals
15:35:55 <n0ano> generic nova will have `lots` of proposals
15:36:15 <bauzas> Ctrl+F is worth it
15:37:22 <n0ano> I just wanted to raise the issue of Atlanta sessions so that's enough for today
15:37:28 <n0ano> #topic opens
15:37:38 <n0ano> anything else anyone wants to raise today?
15:38:45 <n0ano> hearing crickets so I'll thank everyone and we'll talk again next week
15:38:46 * bauzas hearing
15:38:56 <n0ano> tnx all
15:38:58 <bauzas> n0ano: sure thanks
15:39:00 <n0ano> #endmeeting