16:03:53 <rkukura> #startmeeting networking_ml2
16:03:54 <openstack> Meeting started Wed Jan 27 16:03:53 2016 UTC and is due to finish in 60 minutes.  The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:03:55 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:03:57 <openstack> The meeting name has been set to 'networking_ml2'
16:04:02 <rkukura> #topic Agenda
16:04:16 <rkukura> #link https://wiki.openstack.org/wiki/Meetings/ML2#Meeting_January_27.2C_2016
16:04:30 <rkukura> very lite agenda today - anyone have anything to add?
16:05:10 <rkukura> OK, lets breeze through it then
16:05:20 <rkukura> #topic Announcements
16:05:44 <rkukura> Only announcement I have is that the summit session proposal deadline is February 1
16:05:59 <rkukura> #link https://www.openstack.org/summit/austin-2016/call-for-speakers
16:06:10 <scheuran_> hotels seem already to be sold out :(
16:06:34 <Sukhdev> yes, please make bookings quickly -
16:06:45 <Sukhdev> other day somebody mentioned that in Neutron meeting as well
16:06:58 <rkukura> scheuran_: manage to find one?
16:07:31 <scheuran_> rkukura, I need to take one more outside of the center..
16:07:40 <scheuran_> need to check the public transport first...
16:08:11 <rkukura> scheuran_: maybe Uber
16:08:20 <rkukura> any other annoucements?
16:08:35 <scheuran_> but in the openstack foundation proposed ones around the venue there are only some king suites left ...
16:08:59 <rkukura> I should have waited to book ;)
16:09:30 <rkukura> #topic limited-portsec implementation
16:09:48 <rkukura> yalei: Looks like actively being reviewed/updated - Any update on this?
16:10:10 <yalei> yes, it is in reviewing and some update based on comments
16:10:22 <rkukura> Seems on track to make Mitaka then?
16:10:36 <yalei> hope so
16:10:59 <yalei> welcome more comments :)
16:11:30 <yalei> I try to ask kevin to review it again
16:11:52 <rkukura> amotoki said he would also ping kevinbenton about it, right?
16:12:20 <rkukura> OK, anything else on limited-portsec?
16:12:32 <yalei> no, thanks
16:12:59 <rkukura> #topic Open Discussion
16:13:23 <rkukura> scheuran_: Any update on your agent work (forgot to put that in agenda)?
16:14:00 <scheuran_> rkukura, yeah got a +2 from ihrachys - thanks!!! Just need another core approval to get the first patchset in
16:14:19 <rkukura> great!
16:14:26 <scheuran_> rkukura, and after that it's "just" moving the code into a separate file
16:14:31 <neiljerram> As I'm here, and there's time: rkukura, how do you feel about the use of the multiprovider extension that's being discussed in the routed networks spec?
16:14:47 <scheuran_> currently I'm coding the macvtap agent along those new interfaces...will push it up later this week
16:14:58 <scheuran_> that's it
16:15:08 <rkukura> neiljerram: thanks for bringing that up!
16:15:20 <rkukura> scheuran_: sounds good - thanks!
16:15:42 <rkukura> neiljerram: I’m not up to date on that discussion, but will try to catch up today
16:15:54 <rkukura> Can you summarize the proposal?
16:16:12 <neiljerram> Cool, it will be good to have your input.
16:16:42 <rkukura> I did see there was an email thread on monolithic plugin vs. ML2 driver, and I haven’t followed up on that either (or maybe that’s the same discussion)
16:17:18 <neiljerram> To be honest I'm not fully up with the detail myself, but I believe the proposal now is to use multiprovider to meet the use case in that spec
16:17:43 <neiljerram> That other thread was me too, but not closely related.
16:17:51 <rkukura> neiljerram: So would this involve multiple network segments that do not have L2 bridging between them?
16:17:56 <neiljerram> But yes, I'd appreciate your input on that too!
16:18:07 <neiljerram> Yes, that's right.
16:18:33 <neiljerram> The 'large deployers' physical networks consist of L2 segments with routing between them
16:18:45 <rkukura> So tenants would need to be aware that the networks they are creatining are not L2 bridge domains in that case
16:19:14 <neiljerram> Indeed, and that part of the reqt is proposed to be met by adding an 'l2-adjacency' flag to the Network object.
16:19:39 <rkukura> neiljerram: Interesting - I was just going to ask if anything like that was being proposed
16:19:48 <neiljerram> So in the large deployers case, and also for Calico (my project), l2-adjacency would be false.
16:20:05 <rkukura> Is this something the tenant can request, or just see what they get?
16:20:28 <neiljerram> I think we're mostly talking provider networks here.
16:20:57 <rkukura> neiljerram: I see - tenants wouldn’t even be creating their own networks or subnets in this use case, right?
16:21:21 <rkukura> Just creating ports on provider routed networks that don’t have l2-adjacency
16:21:40 <neiljerram> Yes, at least that's the primary target.
16:22:08 <neiljerram> I think it still works, though, to mix provider routed networks with traditional l2 tenant networks.
16:22:12 <rkukura> neiljerram: Seems to make sense.
16:22:30 <Sukhdev> neiljerram : can you post the link to the spec?
16:22:52 <neiljerram> IIUC, using multiprovider is interesting because it helps with the question of knowing which hosts are connected to each segment.
16:22:52 <rkukura> ML2’ers - please give some thought to the idea of relaxing the notion that ML2 networks are always L2 broadcast domains, and comment on these discussions!
16:23:13 <neiljerram> #link https://review.openstack.org/#/c/225384/
16:23:33 <rkukura> neiljerram: Thanks. Also, would you like us to put this on the agenda for next weeks ML2 meeting?
16:23:39 <neiljerram> Thanks very much everyone!
16:24:06 <neiljerram> Yes, great idea.  Also I will try to make sure that I'm fully ramped up on the detail myself, by then!
16:24:31 <rkukura> neiljerram: Sounds good, and feel free to invite others working on this
16:24:51 <rkukura> Any other topics for today’s open discussion?
16:24:51 <neiljerram> sure, I'll invite Carl in particular
16:25:27 <Sukhdev> I have a bit of background on this -
16:26:11 <rkukura> Sukhdev: want to discuss now, or save for next week?
16:26:30 <Sukhdev> I have not looked at the spec, but, I was in the discussion when this topic took place during neutron mid-cycle
16:26:38 <Sukhdev> we can discuss this in the next meeting
16:27:07 <neiljerram> Sukhdev, thanks - but note that the spec has evolved a great deal since then...
16:27:45 <Sukhdev> neiljerram : correct - let me review it to catch up
16:27:56 <rkukura> Ideally, I’d like to see nova/neutron have some more generic way of specifying “the thing to which a port is attached”, where this could be an L2 network, an L3 routed network, or even some policy group kind of thing
16:28:16 <rkukura> But pragmatically, the current proposal looks worth discussing
16:29:05 <rkukura> Anything else on routed networks or anything else today?
16:29:34 <Sukhdev> Any proposal for the summit?
16:29:55 <Sukhdev> How about SG - which we pushed to Newton release?
16:30:13 <rkukura> Sukhdev: Are you asking about summit session proposals, or design sessions?
16:30:39 <Sukhdev> summit session?
16:30:49 <Sukhdev> deadline is next week
16:31:23 <rkukura> I think we have agreement we want to discuss how ML2 can address SG and other extension enforcement as a design session, but I’m not sure anyone is thinking about a main summit session on this, are they?
16:32:33 <Sukhdev> for design session, we have time -
16:33:03 <rkukura> If anyone is looking for main summit session co-presenters or anything, now is a good time to bring that up
16:33:38 <Sukhdev> yup - that was the reason I brought it up
16:33:52 <rkukura> I wish we had something new in ML2 worth presenting, but this cycle has been kind of quiet on that front
16:35:08 <rkukura> Lets see if we can put together some good general ML2 enhancements for Newton, and then we can team up to present in Barcelona!
16:35:45 <rkukura> Anything else today before we wrap up?
16:36:05 <Sukhdev> sounds good
16:36:08 <rkukura> If not…
16:36:09 <Sukhdev> nothing from me
16:36:30 <rkukura> Thanks everyone - have a great week!
16:36:35 <rkukura> #endmeeting