16:01:04 <Sukhdev> #startmeeting networking_ml2 16:01:05 <openstack> Meeting started Wed Jul 8 16:01:04 2015 UTC and is due to finish in 60 minutes. The chair is Sukhdev. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:06 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:08 <openstack> The meeting name has been set to 'networking_ml2' 16:01:32 <Sukhdev> #topic: Agenda 16:01:37 <Sukhdev> #link https://wiki.openstack.org/wiki/Meetings/ML2 16:01:45 <Sukhdev> Good morning folks 16:02:05 <amotoki> hi 16:02:18 <Sukhdev> #topic Announcements 16:02:35 <Sukhdev> I do not have any specific announcement on the agenda 16:02:45 <Sukhdev> does anybody have any announcement? 16:03:03 <shivharis> yes.. hold on 16:03:12 * Sukhdev waiting 16:03:23 <shivharis> the last date for talks is 17th for submissions of talks 16:03:40 <shivharis> am i right? 16:03:52 <Sukhdev> shivharis: for Tokyo summit? 16:03:56 <shivharis> yes 16:04:02 <Sukhdev> shivharis: No it is 15th July 16:04:18 <shivharis> ok, it is fast approaching so best to remind folks 16:04:21 <rkukura> Sukhdev, shivharis: RIght, email says the 15th 16:04:40 <Sukhdev> Good point shivharis - thanks for reminder 16:04:57 <Sukhdev> BTW, I am proposing one talk for the main session - 16:05:06 <Sukhdev> for design summit - we have lots of time 16:05:34 <rkukura> Sukhdev: What’s your topic? 16:06:02 <Sukhdev> rkukura: not finalized yet - but, it will be on the lines of Ironic/Neutron integration 16:06:12 <shivharis> Arvind and I are proposing topology as a service 16:06:37 <Sukhdev> shivharis: I was thinking of that for design session 16:06:52 <Sukhdev> shivharis: I mean for design session for sure 16:07:06 <shivharis> actually the touch points of this is both in nova and neutron 16:07:32 <shivharis> i am getting Nova folks about this 16:08:01 <Sukhdev> shivharis: you should do the same way as i did in last summit - present in design session for both projects 16:08:27 <shivharis> exactly - Arvind and I were thinking that way 16:08:56 <Sukhdev> shivharis: yup - that will be a good idea 16:08:57 <shivharis> but i dont want to lose the opportunity to do a talk in the general sessions 16:09:24 <Sukhdev> shivharis: sure - you should try that too 16:09:25 <asomya> sukhdev: The main track is to get a broader audience to present the scope of the project 16:10:01 <Sukhdev> asomya: correct - if you can get them to approve it 16:10:12 <amotoki> and sometimes it helps developers understand it. 16:10:30 <Sukhdev> asomya: I think this will be good for both sessions 16:10:36 <shivharis> amotoki: agreed, design talks are a must for both Nova and Neutron 16:10:51 <asomya> Sukhdev: Agreed, maybe we can have both? 16:10:53 <shivharis> Sukhdev: +1 16:11:22 <Sukhdev> shivharis asomya: But, first you have to present it to us - here :-):-) 16:11:32 <shivharis> Sukhdev: +1 16:11:40 <shivharis> +2 16:11:43 <amotoki> shivharis: I am not sure what projects are related to your proposal. only nova and neutron? it is out of scope of this meeting though... 16:11:55 <Sukhdev> anyways, I have it on agenda for today's discussion - so, hold on to your seats :-) 16:12:08 <asomya> amotoki: The first iteration is a bit Neutron and nova focused, the scope can be expanded later to storage etc. as required 16:12:27 <shivharis> asomya: +1 16:12:53 <amotoki> asomya: thanks. I just feel so from the word "topology as a service". 16:13:03 <Sukhdev> amotoki: why do think it out of scope - I think in Neutron, this is more applicable to ML2 drivers than anything else 16:13:42 <shivharis> Sukhdev: ML2 driver is a consumer 16:13:52 <asomya> Sukhdev: Like amotoki said, it's a generic topology service that can be consumed by any other service .. neutron/ml2 is just one of them 16:13:53 <amotoki> Sukhdev: "out of scope of the meeting" is confusing... sorry, precisely speaking, the current topic is "announcement". 16:14:28 <amotoki> but topology as a service is not specific to ml2. 16:14:35 <Sukhdev> amotoki: ah ha got it - yes, I have this topic on the agenda 16:15:06 <Sukhdev> speaking of that - lets get back on track to the agenda 16:15:21 <Sukhdev> #topic: Action Items 16:15:47 <Sukhdev> rkukura: was going to work with manishg to finalize the ML2 mid-cycle 16:16:09 <manishg> Sukhdev: I think discussion is going via email with all those interested. 16:16:10 <Sukhdev> I noticed we have lots of discussion on this, but, we need to finalize the details and plan 16:16:40 <Sukhdev> I think we have all participants here - shall we finalize it here now? 16:16:47 <manishg> and looks like next week is out. So likely dates are late Aug or early Sep (due to other conflicts and release schedules). 16:16:57 <manishg> Sukhdev: sure 16:17:17 <Sukhdev> manishg: based upon the emails, it seems like early sept is best - 16:17:32 <manishg> Sukhdev: sounds good to me. how about others? 16:17:48 <Sukhdev> rkukura made a good point that this will not go into liberty anyway - so, might be better to plan a bit better 16:17:59 <rkukura> Early Sept works best for me at this point 16:18:10 <manishg> week of Sep/7 will be a short week. 16:18:46 <Sukhdev> manishg: is that labor day week? 16:18:47 <amotoki> feature freeze is the first week of Sep. 16:18:52 <amotoki> isn't it too late? 16:19:05 <amotoki> it's just a question 16:19:15 <Sukhdev> amotoki: we are thinking that we will push it for M release - 16:19:19 <manishg> Sukhdev: I think so. 16:19:27 <amotoki> got it. 16:19:56 <shivharis> BTW if you are having issues of hosting, i can arrange taht 16:19:58 <Sukhdev> So, shall we do in the week after the labor day week? 16:20:40 <Sukhdev> shivharis: now we have three bids on the table - shall we start the bidding - I do not take less than $100 bills :-) 16:20:52 <shivharis> $0 16:20:53 <Sukhdev> just kidding :-) 16:21:04 <shivharis> jk 16:21:30 <Sukhdev> who wants to take the action to finalize this - ie. who is hosting, where, when, etc….? 16:21:52 <Sukhdev> rkukura: has already put together an ehterpad for this 16:22:05 <Sukhdev> #link: https://etherpad.openstack.org/p/Neutron_ML2_Mid-Cycle_Sprint 16:22:08 <shivharis> I just need a weeks headstart to arrange 16:22:24 <shivharis> the venue 16:22:33 <rkukura> Is everyone OK with early September in the Bay Area? 16:22:46 <Sukhdev> rkukura: yes 16:23:08 <Sukhdev> shivharis: you are third in the line - $500.00 brings on the top of the line :-) 16:23:15 <rkukura> And the idea of doing the work on a branch, hopefully to merge in early M? 16:24:19 <rkukura> I’ll update the etherpad to try to identify a week in early September 16:24:24 <shivharis> rkukura: +1 16:24:49 <rkukura> I’ll list each week and ask people to put their name with Y or N. 16:24:52 <Sukhdev> rkukura: looks like you are volunteering for planing this 16:24:57 <rkukura> yes 16:25:19 <Sukhdev> yay!!! - thanks rkukura - I will assign you an action 16:25:28 <rkukura> I’d like to nail down the dates by next Wednesday’s meeting 16:25:53 <Sukhdev> #action: rkukura to coordinate with rest of the team to finalize the detailed plan for ML2 mid-cycle sprint 16:26:07 <rkukura> So if you are interested in participating, please get your name onto the etherpad and we’ll include you in any emails. 16:26:26 <Sukhdev> anything else on this topic? 16:26:42 <banix> so lets call it late-cycle. It is more accurate and defenitely more cool. 16:26:52 <Sukhdev> rkukura: I have possible one volunteer from Arista who might join as well - 16:26:52 <rkukura> One last detail is which projects in addition to async will be worked on at the mid-cycle 16:26:59 <shivharis> rkukura: actually put ur name as attending + interested and not attending 16:27:23 <shivharis> split of ML2? 16:27:49 <Sukhdev> rkukura: make sure banix is there to head the cheerleading staff 16:28:16 <rkukura> would be a good idea to move to a separate repo before branching if we are going to do that 16:28:23 <Sukhdev> banix: kidding aside you have a good point 16:28:32 <amotoki> "pre-cycle sprint for M**" rather than "mid-cycle" 16:28:50 <Sukhdev> rkukura: I was thinking stackforge - but, that may be too heavy handed 16:29:23 <rkukura> I think a feature branch in the same repo where ML2 lives would be fine. Any reason to do it in stackforge instead? 16:29:28 <shivharis> github for sure maybe not stackforge 16:29:40 <shivharis> rkukura: like ur idea 16:29:58 <rkukura> We want to be able to easily rebase the branch as other changes are merged 16:30:49 <Sukhdev> I only thought of stack forge so that this can be released as a package and anybody who wants to play with it can pull it from pypi 16:31:17 <rkukura> Sukhdev: Worth considering. Lets get back to that at a future meeting 16:31:20 <Sukhdev> but, I am OK either way - do not want to make it too heavy handed thought 16:31:36 <shivharis> pypi can deal with branches, i think 16:32:09 <Sukhdev> shivharis: not sure - you may be right - may be worth looking into this 16:32:30 <Sukhdev> Any ways - lets move on 16:32:54 <Sukhdev> #topic: L2 modular agent 16:33:20 <Sukhdev> banix: want to give us the details - 16:33:44 <Sukhdev> banix: I mean update - 16:33:59 <banix> Sukhdev: I haven’t made any progress; trying to consolidate wrt ungoing enhancement to the agents 16:35:10 <amotoki> I think "modular agent" has several aspects: 16:35:18 <amotoki> modularity for features 16:35:21 <Sukhdev> yamamoto - anything you want to share? 16:35:25 <amotoki> modularify for backends 16:36:13 <Sukhdev> amotoki: correct 16:36:46 <amotoki> re: the first point, ajo proposal is worth considered. 16:37:33 <yamamoto> Sukhdev: ? 16:37:55 <yamamoto> Sukhdev: about modular agent? nothing from me 16:38:09 <Sukhdev> yamamoto: sounds good - 16:39:11 <Sukhdev> anything else on this topic? 16:39:23 * Sukhdev waiting 16:39:45 <Sukhdev> #topic: Physical Topology Discussion 16:39:52 <amotoki> banix: could you update the progress in bug 1468803? 16:39:52 <openstack> bug 1468803 in neutron "Modular L2 Agent" [Undecided,Confirmed] https://launchpad.net/bugs/1468803 16:40:01 <amotoki> or it there any place to see? 16:40:15 <Sukhdev> #undo 16:40:16 <openstack> Removing item from minutes: <ircmeeting.items.Topic object at 0x9556490> 16:40:37 <banix> amotoki: will do 16:40:49 <Sukhdev> banix amotoki : thanks 16:41:03 <Sukhdev> #topic: Physical Topology Discussion 16:41:18 <Sukhdev> shivharis asomya: Folks want to give us some update? 16:41:33 <Sukhdev> got any link to share? about your work so far? 16:41:48 <shivharis> We are in the process of refining the design 16:42:05 <shivharis> we are discussing this with folks in Nova 16:42:14 <Sukhdev> shivharis: care to share what ever you have? 16:42:34 <shivharis> as discussed earlier it will have touch points in both nova and neutron 16:42:43 <shivharis> possibly cinder as we go forward 16:43:12 <amotoki> shivharis: is it realted to scheduler or more? 16:43:19 <shivharis> we will share it in the next week or two and possibly at the early M meeting we are planning 16:43:38 <shivharis> amotoki: right 16:44:11 <shivharis> the input to db model will be getting info of the host ids from nova 16:44:11 <Sukhdev> shivharis asomya: I am a bit lost here about this service 16:44:43 <Sukhdev> shivharis asomya : Are you actually doing any work to tie it with any of these projects or just a white paper? 16:45:00 <shivharis> we will present the entire picture in the pre-M meeting we are planning 16:45:21 <asomya> Sukhdev: A high level overview, it's a generic topology manager that can store and retrieve topology, anything can ask or store topolgy via the API 16:45:41 <Sukhdev> asomya shivharis: there is a prior work done on this topic by issaku - which was presented in Atlanta summit - have you seen it? 16:45:49 <Sukhdev> is it on the same lines? 16:45:51 <yamahata> Sukhdev: asomya what first primary use case do you have in mind? 16:45:52 <shivharis> currently it will be nova and neutron - the design will not exclude cinder 16:46:17 <asomya> yamahata: The primary use case is Neutron and nova 16:46:18 <yamahata> s/Sukhdev/ shivharis / 16:46:57 <yamahata> asomya: VM scheduling? 16:47:06 <shivharis> at a high level there are discovery agents feeding the topology discoverd to a consolidator 16:47:25 <asomya> yamahata: Getting hardware information required by driver and plugins to configure the physical fabric 16:47:42 <amotoki> i think yamahata is asking what is the main target as a high-level goal. what can it achieve? 16:47:43 <shivharis> the consolidator will expose an API that can be queried by any project 16:48:26 <yamahata> asomya: then your first goal is to abstract fabric management? 16:48:51 <yamahata> amotoki: exactly. thanks for explanation. 16:48:52 <shivharis> the main target at a high level to begin with with will be neutron or ML2 16:49:02 <asomya> yamahata: expose it rather than abstract it 16:50:02 <Sukhdev> shivharis asomya : Are you working on a POC, along with a use case, for this or it is just white paper? 16:50:22 <shivharis> POC is underway 16:50:25 <amotoki> i think what asomya means is to export it in a unified/general way 16:50:47 <Sukhdev> shivharis: good 16:51:39 <Sukhdev> shivharis asomya: So, this leads me to the next question - we were planning on working on this during mid-cyle (or late-cyle or pre-cycle) sprint - 16:52:02 <Sukhdev> Do we still need to put on the agenda for the sprint - 16:52:29 <Sukhdev> without having much information about it, it may be hard to dive into the implementation - 16:52:46 <asomya> Sukhdev: This is outside of Neutron, development is independent. we can talk a bit about API refinement etc in the late-cycle 16:54:00 <Sukhdev> asomya: so, this is not really targeted to be implemented in the sprint then, right? just for the discussion - am I getting it right? 16:54:08 <asomya> Sukhdev: correct 16:54:34 <Sukhdev> asomya: thanks - this clears my confusion 16:55:05 <Sukhdev> rkukura: so, you can update the etherpad accordingly so that people have right expoctation 16:55:16 <rkukura> ok 16:55:24 <Sukhdev> anything else on this topic? 16:55:47 <Sukhdev> asomya shivharis: Thanks 16:55:55 <banix> does this relate to metering monitoring efforts? 16:56:05 <banix> already being undertaken 16:56:13 <Sukhdev> banix: no 16:56:22 <banix> I mean there are coceptually similar 16:56:23 <Sukhdev> banix: I do not think so 16:56:36 <banix> s/there/they/ 16:57:18 <Sukhdev> banix: I would say - metering will query information from the known components - physical topology will discover components 16:57:27 <Sukhdev> banix: for a lack of better words 16:57:38 <banix> Sukhdev: i see. thanks. 16:58:16 <Sukhdev> physical topology helps you discover how your DC topology looks like so that you can make placement decisions 16:58:41 <Sukhdev> once the VMs are placed, then metering kicks in to monitor how they are performing 16:58:52 <shivharis> time check, 2 mins 16:58:56 <Sukhdev> banix: hope this helps 16:59:05 <Sukhdev> shivharis: thanks 16:59:11 <banix> well, i have to learn more about the proposed project; i see convergence with discovery/monitoring of virtual resources…. but i am probably off the tracks 16:59:13 <Sukhdev> #topic: Ironic Discussion 16:59:21 <Sukhdev> I will be short and sweet - 16:59:43 <amotoki> banix: perhaps it is about network topolgy about virtual reosurces, ie tenants. 16:59:51 <Sukhdev> For those who do not know - there is a lot of work going on in the Ironic/ML2/Neutron integratipon 17:00:28 <banix> Sukhdev: what’s the best way of tracking that work/ learn about them 17:00:43 <amotoki> Sukhdev is making great work. 17:00:43 <Sukhdev> since we are out of time - let me post the link so that you can discover/read on your own 17:00:56 <Sukhdev> #link: https://wiki.openstack.org/wiki/Meetings/Ironic-neutron 17:01:17 <banix> Sukhdev: thanks 17:01:26 <Sukhdev> banix: the meetings take place every monday at 9am (PT) at #openstack-meeting-4 - please come join us 17:01:49 <Sukhdev> amotoki is one of the active reviewer of the work being done there 17:02:05 <banix> Sukhdev: will try to attend 17:02:29 <Sukhdev> The idea is that ML2 drivers should be able to deal with VMs as well bare metal deployments 17:02:39 <Sukhdev> we are out of time folks 17:02:47 <Sukhdev> thanks for attending the meeting 17:02:54 <Sukhdev> have a wonderful day 17:02:54 <banix> bye 17:02:56 <yamamoto> bye 17:02:59 <Sukhdev> #endmeeting