15:02:00 <mestery> #startmeeting neutron-drivers
15:02:01 <openstack> Meeting started Wed Oct 22 15:02:00 2014 UTC and is due to finish in 60 minutes.  The chair is mestery. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:02 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:03 <regXboi> moo
15:02:04 <openstack> The meeting name has been set to 'neutron_drivers'
15:02:16 <sweston> Is Salvatore in Hawaii? :-)
15:02:19 <mestery> #topic Kilo Summit Schedule
15:02:26 <markmcclain> hi
15:02:31 <mestery> We'll discuss this today, and that's about it since I estimate it will take the full hour
15:02:57 <regXboi> mestery: etherpad link?
15:04:39 <amotoki> #link https://etherpad.openstack.org/p/kilo-neutron-summit-topics-distilled
15:04:40 <mestery> #link https://etherpad.openstack.org/p/kilo-neutron-summit-topics-distilled
15:04:43 <mestery> amotoki: You beat me :)
15:04:59 <mestery> We have 4 slots on Wed, and 7 slots Thur
15:05:04 * regXboi wonders if #undo works
15:05:05 <mestery> So very limited compared to Atlanta
15:05:06 <regXboi> #undo
15:05:21 * regXboi guesses not for non chairs
15:05:32 <mestery> The drivers team has taken all the feedback from the original etherpad and condensed it into what is on the distilled page
15:05:44 <markmcclain> regXboi: yeah non chairs don't get many commands
15:05:47 <mestery> In an attempt to get something manageable
15:06:14 <mestery> The goal with Kilo is paying down technical debt, per my email to the list.
15:06:28 <mestery> And we're trying to focus on things which require F2F time in Paris, not things which can be done on ML or IRC
15:06:32 <mestery> So, that's the back story.
15:06:59 <markmcclain> also we've got the tables on Friday
15:07:12 <mestery> markmcclain: Good point, thanks for reminding me of that.
15:07:19 <markmcclain> so topics of interest to sub groups within the community fit better there
15:07:47 <mestery> Any questions so far from people?
15:07:57 <marun> yes
15:08:14 <marun> I've requested a number of times that we include time for lightning talks in the schedule.
15:08:35 <mestery> You mean like taking one 40 minute slot for 5 minute lightning talks?
15:08:39 <marun> To ensure that those that don't get summit time do get to present to the community briefly about what they want to meet about in the unscheduled times.
15:08:40 <marun> Yes
15:08:57 * mestery thinks on this for a minute
15:09:02 <marun> I think that's only fair given how many people won't be getting time due to the focus on community stuff.
15:09:15 <marun> time -> time in the regular sessions
15:09:20 * regXboi notes: marun: +1
15:09:39 <amotoki> it is useful to form some small groups or BoF.
15:09:56 <markmcclain> so we're very limited on slots
15:10:06 <marun> That's a poor excuse.
15:10:07 <markmcclain> what if we kept a backlog of lightning talk requests
15:10:11 <marun> I think we can spare a single slot.
15:10:19 <markmcclain> and if any sessions run short of their allocation we pull from backlog?
15:10:47 <markmcclain> marun: if the nova/neutron session gets pushed back into our track we'll be in a big time crunch
15:10:49 <mestery> I don't think it's a poor excuse, rather the realities of the situation, but I understand your point of view as well marun.
15:10:50 <marun> While that seems reasonable on its face, I don't think it's wise.
15:10:57 <marun> markmcclain: ^^
15:11:10 <salv-orlando> let’s just look at cost/benefits
15:11:12 <marun> The whole point of lightning talks is that they are scheduled.
15:11:26 <salv-orlando> the cost is 1 slot = 9% of neutron summit time
15:11:28 <marun> Those that are giving them and those that want to attend have a time to be in the same room.
15:11:32 <salv-orlando> what about the benefit?
15:11:56 <salv-orlando> I mean my concernis what benefit there will be in talking 5 minutes about something? Merely community exposure?
15:11:56 <marun> 9% seems a pretty tiny amount to allow the community to focus on what they want to focus on in addition to everything that we consider core.
15:12:08 <marun> Exposure, yes.
15:12:09 <markmcclain> marun: what do you propose we cut?
15:12:39 <marun> markmcclain: good question
15:12:53 <marun> markmcclain: Do you have a priority listing of the sessions scheduled thus far?
15:13:01 <mestery> marun: It's on the etherpad we linked
15:13:02 <marun> markmcclain: Such that you know what would be worth giving up?
15:13:05 <mestery> We've allocated the slots.
15:13:10 <mestery> As best as we could so far.
15:13:14 <marun> You all have been thinking about this an awful lot more than any of us.
15:13:16 <mestery> Thus, something from there needs to go if we were to do this.
15:13:47 <marun> I think my point is that community focus doesn't just include the technical.
15:13:56 <marun> It includes giving people the opportunity to participate.
15:14:04 <mestery> Fair point
15:14:08 <markmcclain> marun: I agree with that sentiment
15:14:23 <marun> So, olive branch.
15:14:46 <marun> There are many that are going to be upset that they don't get time in from of everyone they way they have in the past.
15:15:00 <mestery> Well, no matter what we do, there will be many people who are upset.
15:15:05 <marun> I think most would agree that it's a positive change, because we have lots we have to talk about to ensure our continued success.
15:15:42 <marun> But I don't think it's unreasonable that people still get a limited forum to present why their ideas are important.
15:15:43 <salv-orlando> marun: I think there will be many people wishing to argue on the “continued success” statement ;)
15:16:23 <marun> Anyway, I've spoken my piece.
15:16:40 <mestery> marun: I'm thinking of what we can cut out out now to give a lightning talk slot.
15:16:41 * regXboi notes looking at the list of slots, the only one that might make sense to give up is the last one on the list...
15:17:01 * regXboi notes *MIGHT* not does
15:17:12 <amotoki> a question is how many folks would like lighting talks. Does 8 LT slots fit? who can present?
15:17:20 <marun> I'm not sure why we need a whole session on rpc, frankly
15:17:27 <marun> amotoki: lottery?
15:17:35 <marun> amotoki: the logistics are less important than getting the slot.
15:17:35 <mestery> 6 is more reaslistic
15:17:42 <mestery> 8 is too manyu for 40 mionutes
15:17:45 <marun> mestery: we could do 8
15:17:47 <mestery> marun: Do you agree?
15:17:49 <mestery> marun: How?
15:17:55 <mestery> 40 minutes, time for swapping, 8 is tight, right?
15:17:59 * mestery rhymes his way to reason
15:18:04 <markmcclain> haha
15:18:16 <amotoki> From the driver discussion last week, slots with a bit weeker need is "Plugin/RPC API" (a partof API sessions).
15:18:17 <marun> again, the logistics could be resolved if the commitment is made
15:18:28 <marun> it could be put to a vote, frankly.
15:18:31 <dougwig> 8 only works if you assume no teardown/setup time.  which is unrealistic.
15:18:31 <amotoki> CLI slot might not need a full slot.
15:18:32 <marun> but let's not focus on that.
15:18:39 <mestery> Agreed.
15:18:49 <markmcclain> marun: so folks want exposure for stuff right?
15:19:10 <markmcclain> what if had slides or page of links to ideas and contacts?
15:19:15 <marun> dougwig: they do it at pycon :)
15:19:42 <s3wong> doing 8 means there is absolutely no time for Q&A --- so it is really just getting the 5 minutes exposure (5 minutes pitch)
15:19:48 <marun> markmcclain: There is something special about lightning talks for galvanizing interest in your ideas.
15:20:01 <marun> s3wong: q&a can come after, informally
15:20:02 <markmcclain> marun: that I agree with
15:20:12 <markmcclain> but given time constraints
15:20:15 <mestery> lightning talks have no Q&A
15:20:18 <markmcclain> just concerned
15:20:27 <marun> markmcclain: I think it's better than nothing.
15:20:46 <markmcclain> alternatively we could steal 5 mins of some slots
15:20:52 <marun> but why don't we take this to the list?
15:20:53 <salv-orlando> should we allocate a lightning talk on the tail of each session?
15:20:55 <salv-orlando> or some?
15:21:10 <marun> Ask if people think it's a good idea, and how to do it?
15:21:21 <marun> I mean, the value is really up to the community at large.
15:21:35 <mestery> OK, I think we'll need to allow for some time for lightning talks
15:21:41 <mestery> Logistically, I'll figure out how to make it work.
15:21:45 <mestery> marun: Sound fair?
15:21:52 <s3wong> actually doing 5 minutes on tail end of each session is a good idea, instead of bombarding community with 8 non-stop talks
15:22:09 <marun> s3wong: it's not so bad, ask anyone who's been to pycon
15:22:28 <marun> s3wong: the energy around lightning talks means that they fly by
15:22:32 <marun> s3wong: no chance to get bored
15:23:12 <marun> anyway, mestery has agreed to make it work (thank you!) so we can probably move on.
15:23:22 <s3wong> marun: not bored for sure. wonder if the presentations would stick into people's mind --- but agree we can try
15:23:24 <armax> marun: +1
15:23:25 <dougwig> s3wong: from my experience at atlanta, 40 minutes was either dreadfully too short, or the session in question could've been 15 minutes.  and with some of the contentious slots on the schedule...
15:23:48 <mestery> Lets move on, as marun indicated, we'll slot this in somehow.
15:23:58 <amotoki> +1
15:24:00 <mestery> OK, getting back to what is on the agenda, lets look at Day 1 now.
15:24:48 <mestery> Day has: Dev Process and Procedures, Neutron split, and CLI/Client lib.
15:24:53 <mestery> Nothing contentious in there at least ;)
15:24:58 <regXboi> mestery: well
15:25:39 <mestery> Seems like a solid first day to move forward with. Comments from people?
15:25:41 <regXboi> I *could* argue that CLI/Client Lib goes along with the whole REST/RPC/Plugin API discussion
15:25:51 <mestery> regXboi: You could, sounds like you may? ;)
15:26:00 <regXboi> no, just wondering if I can create a slot
15:26:21 <mestery> regXboi: Heh :)
15:26:35 <regXboi> but otherwise day 1 looks solid
15:26:55 <marun> I'm not sure what we'll have to discuss around client/cli, honestly.
15:27:07 <marun> It's a serious problem, no question.
15:27:17 <markmcclain> marun: imploding them :)
15:27:34 <marun> But most seem to accept its deficiencies and may not have much to say about what needs to be done.
15:27:37 <salv-orlando> the client side discussion is more about reconsidering our approach to what we call client now and discussiing hot it could be reworked
15:27:43 <salv-orlando> regardless of the shape of the API
15:27:51 <marun> It may be worth floating a proposal for refactor on the ML and see if there are any detractors that suggest a session is necessary.
15:27:53 <markmcclain> also we need folks to help align our efforts with the openstack sdk work
15:27:58 <marun> salv-orlando: fair enough.
15:28:06 <markmcclain> the client lib is a consistent complaint we get from users
15:28:10 <salv-orlando> we probably won’t have even time to talk about usability of the CLI
15:28:16 <marun> I guess my point is that the point of the summit sessions now is discussion.
15:28:22 <markmcclain> many still use nova as a proxy because of our client is bad
15:28:38 <marun> That would imply we need a critical mass of interested parties to participate.
15:28:48 <marun> And even though I am one of them, I'm not sure we have that critical mass.
15:28:51 <mestery> That's what summits are for?
15:28:52 <marun> But if you are sure, please ignore me. :)
15:29:58 * marun is sorry for nitpicking
15:30:04 <marun> please continue with day 2
15:30:12 <mestery> marun: Not nit-picking, I appreciate your input here for sure.
15:30:14 <mestery> OK
15:30:21 <mestery> Day 2 (unless there are more comments on Day 1) ...
15:30:26 <amotoki> at least we need to decide the direction of our cli. it may not need a full slot.
15:30:47 <markmcclain> amotoki: the cli is one side of it.. we also have the lib side of it
15:30:50 <mestery> Day 2 starts with a doozie: REST/RPC/Plugin/API
15:30:57 <mestery> Lots to discuss there.
15:31:09 <amotoki> markmcclain: yes.
15:31:15 <salv-orlando> mestery: the last separator should not be a / but a space
15:31:22 <salv-orlando> REST/RPC/Plugin API ;)
15:31:26 <marun> I'm sure of the need for REST and Plugin API.
15:31:26 <mestery> salv-orlando: Ack
15:31:33 <marun> Not so sure about RPC - what exactly is contentious here?
15:31:58 <mestery> Nothing contentious about RPC
15:32:13 <markmcclain> we have design issues to work through
15:32:16 * regXboi wonders if he should ask the async/sync question
15:32:28 * mestery hopes regXboi has had his coffee this morning
15:33:02 <regXboi> so... i'll ask - in this block is there any intent to start looking at async operations or is that already in scope and I've missed it?
15:33:10 * regXboi is thinking towards multi-region support
15:33:14 <markmcclain> regXboi: yes!
15:33:28 <regXboi> markmcclain: :)
15:33:48 <marun> markmcclain: Don't those issues more properly belong in the agent refactor discussion?
15:33:49 <markmcclain> that flows into part of the REST API and the behavioral semantics of the what happens
15:33:56 <markmcclain> marun: that too
15:33:59 <marun> markmcclain: We need a common framework designed around distributed system principles.
15:34:01 <markmcclain> it overlaps with both
15:34:25 <marun> markmcclain: I'd rather see the agent refactor get more time and include discussion around rpc.
15:34:25 <salv-orlando> the funny fact is that most neutron operations are already async. Problem is that we’ve got no indication of that anywhere.
15:34:38 <marun> markmcclain: Or is the goal to provide slack time for rest and plugin api?
15:34:48 <markmcclain> salv-orlando: right
15:35:00 <marun> markmcclain: I think I'd rather see those 2 just get 3 slots rather that pretending that RPC needs our full attention there.
15:35:07 <markmcclain> marun: yes… rpc is fills in a portion of that block
15:35:36 <marun> salv-orlando: async, but not eventually consistent :/
15:35:46 <markmcclain> I would not look at that block as 3 items needing equal time allocation in the 3 slot block… it's really a weighted
15:36:03 <marun> markmcclain: fair enough
15:36:15 <regXboi> honestly, I read it the same way as markmcclain: it's a 120 minute block that can be used as needed
15:36:27 <markmcclain> distribution and we wanted the ability to flow and 3 slots felt like a natural allocation for all 3 items but 1 slot was not enough for some
15:36:33 <mestery> See? Nothing contentious about the first slot on day 2. :)
15:36:37 * mestery ducks
15:36:51 * regXboi hopes mestery was joking :)
15:37:04 <salv-orlando> marun: no need to disturb eventual consistency… suffice to say that there’s no tracking of the operation, expect for what the agent decides to tell us indirectly about it
15:37:15 <marun> salv-orlando: magic? ;)
15:37:33 * regXboi wonders - do we need a slot for adv svc spin out?
15:37:39 <amotoki> i think the slots are to dicsuss what behaviors are expected and visible to REST API, so there are many potential topics including tasks.
15:37:40 * regXboi is reading ahead
15:37:41 <salv-orlando> marun: it is correct to say that we currently rely on the agents being nice and merciful
15:37:53 <markmcclain> regXboi: we have to walk through the mechanics
15:37:54 <mestery> regXboi: Surely you jest?
15:37:55 <markmcclain> and timing
15:38:25 <markmcclain> amotoki: tasks is part L2 agent refactor and part plugin API
15:38:32 <salv-orlando> regXboi: the adv svc spin out poses challenges regarding API design and db management that we need to achieve consensus on.
15:38:35 <regXboi> markmcclain, mestery: wondering what the end result is supposed to be - but I'll ask that when we get there
15:38:50 <mestery> Note that IPAM now has it's own cozy slot in Paris.
15:39:07 <armax> amotoki: what do you mean?
15:39:17 <s3wong> salv-orlando: agreed, I think adv. service spin out is an important item to discuss with some f2f time
15:39:21 <regXboi> mestery: yeah, that one still doesn't quite grok
15:39:24 <carl_baldwin> Did I hear IPAM?
15:39:28 <amotoki> armax: "task" related topic is related to API.
15:39:53 <markmcclain> carl_baldwin: yep.. we're actually going to talk about it
15:39:56 <mestery> carl_baldwin: You did, and pluggable IPAM even
15:40:03 <armax> amotoki: currently it’s being captured in the etherpad for ‘Resource lifecycle and status representation'
15:40:08 <armax> amotoki: but that can be moved
15:40:15 <carl_baldwin> I’ve had IPAM on my mind for a while.
15:40:17 <markmcclain> instead of kick can down the road
15:40:23 <amotoki> it is a topic how we manage states of operations.
15:40:27 <amotoki> armax: thanks
15:40:58 <regXboi> so... I'll come back and ask on the adv svc spin out (since that's the next slot) - the etherpad doesn't reveal what the expected outcomes might be
15:41:25 <mestery> regXboi: I think markmcclain indicated what we expect, to work out some details and finalize a roadmap for this in Kilo.
15:41:56 <regXboi> a roadmap for Kilo or a roadmap for future releases starting with Kilo?
15:41:58 <armax> regXboi: I guess the outcome is to be determined at the session
15:42:33 <s3wong> mestery: regXboi made a great point --- for all these selected design sessions, do we have designated people to prepare for presentation?
15:42:42 * regXboi is trying to nail down the ambiguity
15:42:50 <armax> regXboi: we can start brainstorm on the etherpad for sure
15:43:16 * regXboi wonders where he made the point ascribed to him
15:43:21 <mestery> s3wong: Yes for most, for others we're working on this.
15:43:25 <salv-orlando> s3wong: we might designate session chairs but I don’t think we need presentations or slides
15:43:30 <mestery> regXboi: Me too actually, I don't know how s3wong made that correlation :)
15:43:39 * mestery thinks we should ban slides again
15:43:48 <mestery> Because ... PPT
15:44:08 <salv-orlando> or keynote for that matter, or even openoffice
15:44:08 <regXboi> so... I'm still trying to nail down an ambiguity - a roadmap for Kilo or a roadmap for future releases starting with Kilo?
15:44:19 <salv-orlando> it’s just the idea of the “talk” that should be banned
15:44:23 <mestery> regXboi: Roadmpa for adv svcs spinout?
15:44:27 <mestery> salv-orlando: +1000
15:44:28 <regXboi> mestery: yes
15:44:30 <markmcclain> mestery: +100000000000 for banning PPT
15:44:34 <anteaya> s3wong: I think what neutron is trying to do is move away from presentation format, which other programs don't use, and move to discussion format, which is rather the point of the summit
15:44:47 <mestery> anteaya: +1000000000
15:45:00 <mestery> regXboi: The roadmap for both, kilo and the spinout
15:45:04 <s3wong> regXboi: when you said you didn't know what to discuss in adv. service spinout session --- it implies there has to be people responsible for the actual discussion point --- that's how I perceive it :-)
15:45:28 <markmcclain> right… we'll someone to facilitate discussion
15:45:36 <salv-orlando> s3wong: that would setup in an etherpad collaboratively compiled before the summit
15:45:39 <markmcclain> and probably have a strawman proposal to work from
15:45:44 <s3wong> salv-orlando: sounds good
15:45:47 * regXboi hopes rooms have whiteboards or easels for unconference type things
15:45:51 <salv-orlando> I think we already did is for Atlanta, didn’t we
15:45:52 <mestery> Less leaders, more collaboration?
15:46:04 <markmcclain> it should be noted that the spinout depends on the results of the discussion from day 1
15:46:08 * salv-orlando is a follower looking for a leader
15:46:10 <markmcclain> so would expect changes
15:46:22 <mestery> Yes, some of this is dynamic, as these types of things tend to be
15:47:11 <regXboi> hmm, ok I see my confusion... in my head, I'm already thinking of adv svcs as described in the etherpad as an 'M" thing
15:48:11 <dougwig> M?  as in release M?
15:48:13 <mestery> regXboi: Moar coffee
15:48:30 * regXboi doesn't drink coffee
15:48:36 <s3wong> salv-orlando: in design summit session? no --- we didn't use whiteboard (just etherpad); in pod area, yes
15:48:43 * regXboi notes nobody wants to see him hopped up on caffeine
15:49:10 <mestery> Lets not confuse adv. svcs (LBaaS, VPNaaS, and FWaaS) with service chaining here, which is I think what regXboi was doing.
15:49:44 <regXboi> mestery: not precisely
15:49:46 <s3wong> particularly when we talk about "spinning out" --- clearly it is in context of the existing advanced services (LB/FW/VPN)
15:49:53 <mestery> s3wong: Yes
15:50:05 <regXboi> um
15:50:23 <regXboi> mestery: that etherpad needs updating because that context is *NOT* clear
15:50:32 <mestery> regXboi: Please feel free to update. :)
15:50:46 <mestery> Time to move on: Eveyrone is ok talking IPAM? /cc carl_baldwin
15:50:47 <mestery> ???
15:51:10 <carl_baldwin> I am
15:51:23 <salv-orlando> Or anybody NOT OK talking about IPAM?
15:51:30 <mestery> salv-orlando: That's a beter way to discuss this I think.
15:51:38 <mestery> Or frame it, rather
15:51:41 * salv-orlando if you’re one of those I’ll assign you all the deadlock and lock wait timeout bugs that come out from ipam
15:51:52 <mestery> OK, IPAM it is. Moving on.
15:52:01 <mestery> Refactoring in agents and paying down technical debt there.
15:52:10 <mestery> This is an area which overlaps a few things.
15:53:41 <mestery> And finally, we have a slot for retrospective.
15:54:04 <mestery> I'm personally of the opinion that this is ok, but that it would be better to do this post-Summit once people ahve recovered from partying (and design summit sessions).
15:54:25 <regXboi> mestery: that was the slot I was thinking of suggesting for LTs
15:54:32 <s3wong> mestery: that is a prime candidate to be replaced by lighting talks?
15:54:32 <mestery> regXboi: Me as well to be honest
15:55:01 <mestery> I think the first Neutron meeting post-Summit could be used entirally for priorities and we could nail them down there.
15:55:17 * mestery notes 5 minutes left ...
15:55:57 <amotoki> don't we need to discuss about Friday?
15:56:14 <mestery> amotoki: Yes, but really we can allow signups there as we have 2 3.5 hour blocks
15:56:23 <mestery> At a round table in a room with other round tables and discussions going on
15:56:35 <mestery> So for that, we can push things which require a smaller # of people or things which overflow.
15:56:41 <mestery> Just my current thinking
15:57:22 <markmcclain> there will also be the pods as in past years too
15:57:36 <mestery> OK, since we're almost out of time, I think we can wind this down now.
15:57:42 <markmcclain> so I think that can help folks advertise when pods sessions are occuring too
15:57:47 <mestery> I will finalize the schedule and upload to sched.org by Friday.
15:57:57 <mestery> So please reach out to me if you have questions or concerns before then.
15:58:02 <markmcclain> since in many instances lightning talks are really just open to recruit more for wider discussion
15:58:07 <mestery> And thanks for attending this week everyone!
15:58:23 <mestery> #endmeeting