20:00:13 <asalkeld> #startmeeting Heat
20:00:14 <openstack> Meeting started Wed Oct 22 20:00:13 2014 UTC and is due to finish in 60 minutes.  The chair is asalkeld. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:00:17 <openstack> The meeting name has been set to 'heat'
20:00:27 <asalkeld> #topic rollcall
20:00:34 <asalkeld> o/
20:00:34 <stevebaker> \o
20:00:35 <shardy> o/
20:00:41 <BillArnold> o/
20:00:41 <skraynev> \o/
20:00:49 <jpeeler> o/
20:00:56 <rpothier> o/
20:01:01 <tango> o/
20:01:15 <asalkeld> #link https://wiki.openstack.org/wiki/Meetings/HeatAgenda
20:01:45 <pas-ha> o/
20:01:55 <asalkeld> #topic review actions
20:02:11 <asalkeld> #link http://eavesdrop.openstack.org/meetings/heat/2014/heat.2014-10-15-12.00.html
20:02:13 <stevebaker> zaneb: ping
20:02:20 <zaneb> o hai
20:02:53 <asalkeld> so the only action was to review the autoscaling restructure
20:03:03 <asalkeld> hopefully people did that
20:03:13 <zaneb> I think I did that
20:03:26 <asalkeld> #topic add topics
20:03:35 <tspatzier> hi all, sorry I'm late
20:03:51 <asalkeld> hi tspatzier
20:04:16 * ryansb missed rollcall too
20:04:27 <asalkeld> now we see you ryansb
20:04:39 <asalkeld> I'll move along
20:04:58 <asalkeld> #topic summit sessions
20:05:04 <asalkeld> #link https://etherpad.openstack.org/p/kilo-heat-summit-topics
20:05:32 <asalkeld> the idea is to vote on them
20:05:51 <skraynev> asalkeld: I try to do it :)
20:05:53 <asalkeld> unless people think we need to merge/break them up
20:06:12 <asalkeld> so let's do "upgrades"
20:06:29 <asalkeld> (first topic)
20:06:33 <asalkeld> thanks skraynev
20:07:17 <zaneb> asalkeld: can you clarify #3? "API" can mean a lot of things
20:07:28 <asalkeld> python api
20:07:37 <zaneb> plugin API?
20:07:38 <asalkeld> what does resourcev2.py
20:07:46 <asalkeld> look like
20:07:55 <asalkeld> zaneb, yea
20:08:04 <shardy> Originally in the convergence specs we agreed we weren't going to break the existing plugin interface, I guess that is no longer the case?
20:08:15 <shardy> maybe I missed that conclusion from the meetup
20:09:00 <asalkeld> shardy, ok - i just don't see how
20:09:04 <therve> shardy, The interface is somewhat implicit given the Resource inheritance
20:09:14 <therve> eg can we keep self.stack?
20:09:42 <asalkeld> therve, we need to write some code I think to figure this out
20:10:06 <asalkeld> the bigger issue is do we allocate "convergence" some time
20:10:08 <therve> Probably
20:10:16 <shardy> asalkeld: Ok, cool, I guess I'm just wary of using a whole session to bikeshed on interfaces without that PoC code
20:10:23 <therve> shardy, +1
20:10:32 <asalkeld> do we need sessions for convergence or punt to friday
20:10:33 <zaneb> asalkeld: I'm writing some code at the moment
20:10:38 <asalkeld> zaneb, nice
20:10:46 <shardy> but if it has to change, then I guess it's something important to discuss
20:10:55 <zaneb> tentative conclusion: new resource plugin API could be stage 2
20:10:58 <asalkeld> zaneb, feel free to totally propose new sessions
20:11:11 <asalkeld> (but fast)
20:12:59 <therve> shardy, I think the hypothesis at the meetup was "try to keep compatibility, but makes optional stuff slow if needed"
20:13:03 <stevebaker> should we have one convergence session, then continue on Friday?
20:13:25 <asalkeld> stevebaker, i think we should have one or two even
20:13:34 <stevebaker> then structure that one session so that it is as constructive as possible
20:13:35 <asalkeld> just to "warm" up
20:13:44 <stevebaker> or 2, sure
20:13:44 <shardy> therve: Ok, +1 on that
20:14:01 <asalkeld> i see zaneb is proposing now:( - naughty
20:14:07 <asalkeld> really late zaneb
20:14:23 <zaneb> <asalkeld> zaneb, feel free to totally propose new sessions
20:14:31 <shardy> if we have to have a degraded performance compatibility shim, that's fine, it's just that I know there are folks with out of tree stuff who won't be happy if we break the world for them ;)
20:14:47 <asalkeld> zaneb, i saw autoscaling
20:14:47 <zaneb> asalkeld: this is based on stuff I have learned while prototyping this week
20:15:00 <asalkeld> ok
20:15:18 <asalkeld> zaneb, can you push your poc somewhere?
20:15:36 <zaneb> i.e. stage 1 of convergence can probably enable autoscaling to do the stuff RAX wants in Juno
20:15:44 <zaneb> asalkeld: yes, will do
20:16:01 <BillArnold> zaneb that will include a use case for rapid autoscaling updates?
20:16:11 <zaneb> asalkeld: it will be here: https://github.com/zaneb/heat-convergence-prototype (close to pushing, but keep finding new things to try out ;)
20:16:27 <zaneb> BillArnold: yes, that's what I mean
20:16:41 <asalkeld> i can just make "convergence #1" and "convergence #2" topics
20:17:00 <asalkeld> put what we want there
20:18:22 <asalkeld> another question: since we have an audience do we want to focus more on the usablity stuff?
20:18:30 <asalkeld> (maybe 2 sessions)
20:19:17 <shardy> asalkeld: IMO one session and spill over into friday
20:19:29 <asalkeld> ok
20:19:37 <stevebaker> yeah
20:19:38 <skraynev> it dependences on audience
20:19:41 <shardy> It's super important, but then so is most of the other stuff
20:19:57 <skraynev> one session may be without any feedback
20:20:10 <stevebaker> the formal session to get all the issues on the table, and Friday to cover planning to fix them
20:20:19 <skraynev> possibly solution with Friday will be better
20:20:36 <zaneb> shardy: we should optimise the scheduled sessions for users and other projects to participate
20:20:47 <therve> Wait since when we fix issues in sessions </joke>
20:20:49 <skraynev> +1 stevebaker
20:21:44 <stevebaker> therve: I said _plan_ to fix them. We can then ignore the plan ;)
20:21:52 <asalkeld> lol
20:22:14 <skraynev> therve: 1 session to implement all what we want, 2 - to understand what we did, two other to fix it all ;)
20:22:15 <asalkeld> skraynev, do you want to explain the use case for partial stack
20:22:38 <stevebaker> skraynev: that sounds like a topic for the beer track
20:22:49 <asalkeld> har
20:22:56 <skraynev> I have not nothing to add for use cases in etherpad
20:23:02 <asalkeld> "red, red wine"
20:23:10 <shardy> zaneb: Agreed, I was just questioning if two sessions on one topic (other than convergence) was justified
20:23:20 <skraynev> asalkeld: oh. I agree with wine
20:23:49 <stevebaker> shardy: although usability is an enourmous topic
20:24:07 <skraynev> stevebaker: why not? :) I can combine these things. beer + use case :)
20:24:22 <stevebaker> also should we arrange the chairs in a proper circle and ban the microphone?
20:24:30 <shardy> stevebaker: Yeah, true
20:24:48 <zaneb> shardy: we actually have a lot of time though. we're one of the few projects not to lose any sessions, and we gain a whole day of meetup
20:24:50 <stevebaker> the mic is good for comedy, but it does get in the way a bit
20:25:11 <ryansb> stevebaker: the plus side of a mic is it helps remind people not to talk over each other.
20:25:35 <zaneb> I still don't know how to feel about the mics :/
20:25:42 <stevebaker> ryansb: BLAH BLAH BLAH did you say something?
20:25:55 <shardy> zaneb: Sure, I just saw 11 proposals and 5 slots available ;)
20:26:00 <skraynev> stevebaker :)
20:26:05 <skraynev> good example
20:26:12 <ryansb> stevebaker: nope, must've been the wind.
20:26:13 <asalkeld> i am going to kill the convergance #2 and overwrite with zanes
20:26:13 <zaneb> shardy: I thought we had 7 or 8?
20:26:14 <skraynev> explain a lot of ...
20:26:37 <shardy> zaneb: doh!  I was only looking at Wednesday
20:26:46 <shardy> I assumed it was all on one day like last time
20:26:48 <shardy> cool :)
20:27:40 <asalkeld> can you all recheck you votes for convergence #2
20:28:04 <skraynev> asalkeld: probably email with end list of chosen  sessions will be nice :)
20:28:32 <asalkeld> convergence #3 gone
20:28:35 <zaneb> asalkeld: that's not really convergence imho. it's something that we've been planning for ever that will be unblocked by convergence
20:28:45 <zaneb> (autoscaling, that is)
20:28:53 <asalkeld> mmm
20:29:07 <asalkeld> zaneb, add some details please
20:32:17 <asalkeld> ok got 5 in the boat
20:33:08 <stevebaker> there are a lot of drugs being consumed in https://etherpad.openstack.org/p/heat-useablity-improvements
20:33:19 <asalkeld> *what*
20:33:43 <asalkeld> have things gone to the weeks
20:33:46 <asalkeld> have things gone to the weeds
20:34:41 <stevebaker> its fine, just a lot of people desribing their problems by suggesting solutions
20:34:49 <asalkeld> ok we are done to the last session
20:35:19 <shardy> stevebaker: It is a bit of a soup of ideas, but it's a starting point, at least
20:35:20 <asalkeld> maybe zaqar and spillover (if any) from usablity
20:35:23 <stevebaker> asalkeld: we could make that a floating session for now?
20:35:33 <asalkeld> what's that mean
20:35:39 <asalkeld> "floating"
20:35:57 <asalkeld> stevebaker, ^
20:36:11 <stevebaker> decide on the day, for any other session that really needs more attention. We've done it at least once before
20:36:26 <stevebaker> although that is Friday
20:36:28 <asalkeld> ok, maybe the last one
20:36:42 <asalkeld> good idea stevebaker
20:36:49 <asalkeld> I'll do that
20:37:01 <asalkeld> and maybe cover some of these others
20:37:33 <zaneb> asalkeld: I don't see any convergence sessions on the list?
20:37:56 <asalkeld> that was your one:-O
20:38:11 <asalkeld> i guess it's gone, I'll reshuffle
20:38:13 <zaneb> mine was autoscaling
20:38:15 <stevebaker> oh, we need to discuss documentation at some point. We could do it during usability but then it really would need 2 sessions
20:38:30 <zaneb> if I have to choose I will take convergence over autoscaling
20:38:31 <skraynev> shardy: I added example for partially successfully creation :)
20:38:36 <shardy> I'm confused, is "Autoscaling API" the convergence session now?
20:38:47 <shardy> That seems to me like a separate, but related, topic
20:38:56 <zaneb> shardy: I think asalkeld was thinking that, but it shouldn't be
20:38:57 <asalkeld> zaneb, there was my one with the restarting actions
20:39:15 <shardy> Getting clarity on the distributed worker/observer model seems a precursor to me
20:40:00 <shardy> zaneb: Ok, as long as we aren't going to burn too much time on the "API" aspects vs the the underlying architecture
20:40:14 <zaneb> asalkeld: tbh I'd be fine with having one big "convergence" bucket and we fill in what we want to put in it later
20:40:30 <asalkeld> ok, i am fine with that too
20:40:38 <zaneb> shardy: totally agree, the actual ReST API is the least interesting part
20:40:44 <stevebaker> zaneb: as long as it gets structured before the session, otherwise we might just ramble
20:41:11 <zaneb> stevebaker: +1
20:41:24 <shardy> Seems like we have two parts - convergence #1 is distributed workers and polling (make heat work at scale), #2 is continuous convergence autoscaling/healing etc
20:41:29 <zaneb> asalkeld: was it you who created "convergence #1 - All about the graph"?
20:41:39 <asalkeld> yeah,
20:41:42 <zaneb> ok
20:41:53 <asalkeld> you were suposed to fill that out
20:42:00 * zaneb can come prepared to talk about that
20:42:35 <shardy> we've had folks constantly talking about #2 lately, but we've not even started #1, so I'd rather start there
20:42:46 <zaneb> shardy: I'm coming to the opinion that #1 is distributed workers and autohealing and #2 is polling
20:43:08 <zaneb> or maybe #2 is autohealing and #3 is polling
20:43:13 <zaneb> polling is last
20:43:34 <stevebaker> zaneb: by polling do you mean the observer service?
20:43:34 <zaneb> I will email the list about it this week
20:43:38 <zaneb> stevebaker: yes
20:44:03 <shardy> zaneb: Ok, it will be great to see/hear what you've been looking at
20:44:04 <stevebaker> I have been wondering why observer was first, apart from being the "easy" bit
20:44:12 * zaneb realises that may not have been what shardy meant
20:44:28 <zaneb> stevebaker: it's not the easy bit at all, because it requires new resource plugins
20:45:37 <stevebaker> and the only benefit is offloading some work to another worker
20:46:21 <asalkeld> ok, we have 7 - no spare one
20:46:27 <asalkeld> job done
20:46:30 <asalkeld> :-)
20:46:36 <zaneb> stevebaker: the biggest benefit is that the system becomes much more robust against a worker dying
20:47:01 <asalkeld> zaneb, i want to see the taking over of jobs tho'
20:47:14 * asalkeld hopeful this can be done
20:48:00 <asalkeld> is everyone happy with what has been chosen?
20:48:26 <asalkeld> or is there anything that should be swapped around
20:48:54 <zaneb> tbh I am -1 on nova-network stuff
20:48:59 <zaneb> by all means work on it
20:49:05 <stevebaker> Improve non-neutron (nova-network) resources <- title says nova-network, but all the items are about AWS resources
20:49:26 <zaneb> but I'm not really that interested in a summit session, because it's going away very soon
20:49:42 <asalkeld> zaneb, huh?
20:49:56 <stevebaker> its always 6 months away from going away
20:50:05 <asalkeld> it's about heat, not that it might go away
20:50:11 <asalkeld> who cares
20:50:28 <asalkeld> if it "goes away", do you care about discussing heat
20:50:29 <pas-ha> stevebaker, probably we should retitle it to Nova-network integration
20:51:06 <stevebaker> pas-ha: it should be retitled AWS resource improvements, looking at the items
20:51:44 <stevebaker> pas-ha: I think the only nova-network resource we're missing is OS::Nova::SecurityGroup
20:52:00 <asalkeld> stevebaker, it's about aws as that is the only why you can use non-nova
20:52:11 <asalkeld> stevebaker, loadbalancer
20:52:25 <pas-ha> Nova::Network? Nova::CloudPipe?
20:52:32 <zaneb> asalkeld: I care about stuff that will affect us into the future. stuff that I never have to see again after 1 year... implement anything you like
20:53:05 <zaneb> there's a natural limit to how much damage can be done ;)
20:53:16 <pas-ha> zaneb, I honestly doubt nova-network would be over that soon :/
20:53:34 <asalkeld> zaneb, there is a bigger question
20:53:59 <stevebaker> pas-ha: sure, we need OS::Nova::AllTheThings. I'm not sure a session is needed for that though, just blueprints and code
20:54:00 <asalkeld> will heat move to a differernt release schedule and have to deal with older clounds
20:54:19 <zaneb> stevebaker: +1
20:54:23 <shardy> pas-ha: I guess the question is, do we need a summit session, or is it some work which can just be done
20:54:30 <asalkeld> stevebaker, what would you swap in?
20:54:39 <asalkeld> what's the competitor?
20:54:47 <stevebaker> asalkeld: usability 2?
20:55:03 <pas-ha> I might somewhat agree, we can discuss it in less formal setting
20:55:10 <asalkeld> ok, let me think about that
20:55:30 <asalkeld> there is a *lot* in that usablity etherpad
20:55:32 <stevebaker> pas-ha: don't get me wrong, I'd really like these resources and would be happy to discuss
20:56:14 <asalkeld> pas-ha, are there any "contentious" issues you need to talk about?
20:56:25 <asalkeld> non-obvious stuff
20:56:28 <stevebaker> asalkeld: usability 1 could all be about template format, usability 2 could be all about northbound and southbound APIs, and the things that consume them (heatclient, software config hooks)
20:56:52 <asalkeld> pas-ha or this more about raise awareness of what you want to do
20:57:08 <asalkeld> mmm, time running out
20:57:08 <stevebaker> asalkeld: and usability one could also be about hot-guide, because a lot of the improvements is just needing to document how to do the things
20:57:25 <asalkeld> stevebaker, pas-ha we can move to #heat
20:57:25 <pas-ha> not anything that important, it's just about what I was mainly stumbling recently
20:57:28 <ryansb> (3 minute warning)
20:57:37 <stevebaker> asalkeld: then we could maybe get some docs folk into that session
20:57:38 <asalkeld> doesn't have to be done now
20:57:42 <pas-ha> and to raise awareness, sure
20:57:46 <asalkeld> ok stevebaker
20:57:56 <asalkeld> i'll look at reworking that
20:58:03 <stevebaker> asalkeld: cool, thanks
20:58:08 <asalkeld> stevebaker, you want to make a new session
20:58:12 <stevebaker> ok
20:58:15 <asalkeld> you do one, i can do the other
20:58:26 <pas-ha> so I would choose more usability over nova-network topic
20:58:36 <skraynev> asalkeld: infinite number of sessions...
20:58:38 <pas-ha> if that's the choice
20:58:51 <asalkeld> thanks pas-ha
20:59:08 <asalkeld> maybe we can spend a bit of time on friday on it
20:59:25 <pas-ha> that would be neat :)
20:59:29 <asalkeld> that's all folks
20:59:34 <asalkeld> running out of time
20:59:47 <asalkeld> super productive, thanks
20:59:53 <asalkeld> #endmeeting