17:01:24 <Sukhdev> #startmeeting networking_l2gw 17:01:25 <openstack> Meeting started Mon Sep 14 17:01:24 2015 UTC and is due to finish in 60 minutes. The chair is Sukhdev. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:01:26 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:01:28 <openstack> The meeting name has been set to 'networking_l2gw' 17:01:38 <Sukhdev> hi vikas 17:02:04 <Sukhdev> mondays are crazy for me - back-to-back meetings :-) 17:02:28 <vikas> :-) 17:02:28 <Sukhdev> who else is here? please give me a shout 17:02:36 <Sukhdev> armax: are you here? 17:02:50 <armax> maybe 17:03:28 <Sukhdev> armax: ah ha - I was looking for you in neuron channel - 17:03:43 <Sukhdev> armax: we need you here - as I have couple of items where we need you 17:03:55 <armax> k 17:04:12 <Sukhdev> looking maruti is not here yet 17:04:26 <Sukhdev> #topic: Agenda 17:04:37 <Sukhdev> #link: https://wiki.openstack.org/wiki/Meetings/L2Gateway 17:04:52 <Sukhdev> #topic: Announcements 17:05:15 <Sukhdev> Liberty is around the corner - RC1 is in the works 17:05:30 <Sukhdev> Later we will discuss when do we want to release? 17:05:58 <Sukhdev> Anybody has any announcements? 17:06:22 <gsagie_> I have a general question, is L2GW aimed at solving spanning networks across clouds (public+private) ? 17:06:31 <gsagie_> or is that not an "official" mission 17:07:05 <Sukhdev> gsagie_: thanks and welcome to our meeting 17:08:08 <Sukhdev> gsagie_: I am sure armax will chime in - but, the charter of L2 GW is to provide inter-connect between two L2 broadcast domains 17:08:24 <Sukhdev> to create a one larger L2 broadcast domain 17:08:35 <armax> gsagie_: are you thinking of spanning an l2 broadcast domain over WAN? 17:08:44 <Sukhdev> gsagie_: so, based upon that definition, yes - it covers your use case 17:08:46 <gsagie_> armax: yes 17:09:20 <Sukhdev> gsagie_: so couple of things 17:09:20 <armax> gsagie_: I am sure there are use cases 17:09:31 <armax> gsagie_: I would never use it if it was available though 17:09:35 <armax> performance would suck 17:10:40 <Sukhdev> gsagie_: the two things that I was going to mention is - 1) 17:11:03 <Sukhdev> 1) present implementation is based upon OVSDB - 17:11:04 <armax> can’t say that latency would be enough to dismiss it entirely 17:11:36 <Sukhdev> 2) the orchestration - 17:12:19 <gsagie_> armax: there are some use cases for that in the telco world, but it really depends how you use it and how you implement it, i am not sure actually sending ARP's across WAN make sense but was asking from an API level 17:12:31 <Sukhdev> armax: you mean performance because of encap/decap on both ends? 17:13:38 <Sukhdev> gsagie_: my point 2) was going to cover the API level orchestration 17:13:41 <armax> Sukhdev: not just that alone 17:14:23 <armax> gsagie_: I think that there’s more to understand before we can model it through the api 17:15:27 <armax> we are really crossing the vpn site-to-site use case too 17:15:31 <armax> to some extent 17:16:06 <gsagie_> armax: i agree, i was looking at the current API and think it needs more work for that use case, but didn't want to jump in with it and spend all your time without understanding if thats a project goal 17:18:00 <armax> we’d have to understand more about your use case 17:18:17 <armax> without further details it’s diffucult to assess that it can be a project goal 17:18:24 <armax> in lack of further input I’d say, no 17:18:25 <Sukhdev> gsagie_: even though we have specifically have not said that we will use L2GW for that purpose, but, your use case definitely falls into the broader category 17:18:56 <armax> for instance, are you thinking of bridging another l2 neutron doamin with another neutron one? 17:19:06 <armax> are you thinking of VM to VM traffic? 17:21:07 <armax> we can surely discuss, but it seems premature to expand the project further when the basics are still missing 17:21:14 <gsagie_> One of the use cases is to have a neutron network span across 2 clouds, so you could have VM's deployed in both clouds and all in the same L2 network, 17:21:35 <gsagie_> armax: ok, maybe i will consult with some people and propose a document that describe the use cases 17:22:03 <armax> gsagie_: two separate neutron deployments? 17:22:12 <gsagie_> armax: yes 17:22:26 <armax> gsagie_: have you thought on how you’re gonna handle ipam? 17:22:32 <Sukhdev> thinking in a very simplistic way, one could create an L2 GW on one l2 neutron domain to map VxLAN to a VLAN and create another one in the other neutron domain and use the same VLAN 17:22:53 <Sukhdev> to create connectivity - hence, understanding the orchestration is important 17:23:58 <Sukhdev> gsagie_: I think documenting the use case will really help - 17:24:22 <gsagie_> armax: yes, in our case we have a centralised control plan that does the allocating 17:24:44 <armax> gsagie_: so it’s not two separate neutron deployments then? 17:25:27 <armax> if they share ipam 17:25:29 <armax> I mean 17:26:26 <gsagie_> armax : its two deployment that are controlled from a centralised location, i guess they are not "separate" 17:27:12 <armax> gsagie_: I am sorry, I am not sure what this means 17:29:03 <gsagie_> armax: We have a project (Tricircle) in stack forge, it takes a centralised OpenStack layer and it controls other OpenStack clouds, it has its own implementation (for nova for neutron and so on) that what it does in a nutshell is split the OpenStack API calls and sync the resources between the different clouds 17:29:06 <Sukhdev> gsagie_: I think this complicates - if you keep two separate deployments and use L2 GW on both sides to bridge on the same VLANs, then I could see it working - otherwise, I am at loss as well :-):-) 17:30:19 <gsagie_> and we use availability zone to mark the specific cloud, so the user can using OpenStack API (or UI) to deploy VM's in different clouds by picking different availability zone 17:30:38 <gsagie_> and the project layer, knows to sync all the dependent resources 17:31:13 <gsagie_> armax: https://wiki.openstack.org/wiki/Tricircle 17:31:16 <gsagie_> more information 17:31:17 <armax> gsagie_: sounds like a lot of complication 17:32:00 <Sukhdev> gsagie_: Interesting - this gives me something to read about - 17:32:43 <armax> you could bridge the clouds completely transparently without ever letting the admin know about it 17:32:48 <Sukhdev> gsagie_: coming back to L2GW, I think, if you can document as to how do you envision using it, then we can discuss further and provide guidance 17:33:53 <gsagie_> Sukhdev: ok thanks, i will do that, we have some ideas API/implementation wise 17:34:03 <Sukhdev> armax: that is what I was thinking in my earlier statement - but, it will be good to understand the deployment model better 17:34:14 <armax> Sukhdev: yes 17:35:24 <Sukhdev> #action: gsagie_ to document inter-cloud connectivity use case 17:35:32 <Sukhdev> Folks, I want to return back to the agenda - as we have couple of critical items to discuss 17:35:34 <gsagie_> armax: the tenant has the option to deploy the VM's in which ever cloud he/she wants (or allowed), and of course i guess the pricing/performance model changes accordingly so i am not sure what do you mean "bridge the clouds completely transparently" ? 17:36:15 <gsagie_> ok, sorry for taking too much time 17:36:24 <gsagie_> will document it and post the link, thanks for the time 17:36:33 <Sukhdev> gsagie_: when you write it up, I can help you clarify as to how this will all play out 17:37:27 <Sukhdev> gsagie_: lets discuss the details after you have your thoughts on the document 17:37:39 * Sukhdev returning back to the agenda now 17:37:41 <gsagie_> Sukhdev: ok, thanks 17:37:54 <Sukhdev> #topic: L2GW release planning for Liberty 17:38:42 <Sukhdev> armax: we are close in terms of our work for L2 GW - barring some pending patches (next agenda item) 17:39:30 <Sukhdev> armax vikas: Can you think of things that we need to do before the release? 17:40:06 * Sukhdev waiting 17:40:36 <armax> Sukhdev: what things? 17:41:02 <Sukhdev> any logistical issues - other than merging pending patches 17:41:36 <armax> k 17:41:53 <Sukhdev> armax: by logistics - I mean any release announcement - adding this to neutron Liberty release notes, documentation for neutron etc.... 17:43:11 <Sukhdev> armax: any thoughts? 17:43:30 <armax> none 17:44:23 <Sukhdev> armax: if you think of any, which we need to prepare, please let me know (ping me) 17:44:56 <armax> will do 17:45:43 <Sukhdev> armax: https://review.openstack.org/#/c/202495/ 17:46:09 <Sukhdev> armax: you have -2 on this - have you had a chance to look into the issue? 17:46:22 <Sukhdev> We have tested this and it looks ready to go - 17:46:57 <armax> Sukhdev: ok let me double check with henry 17:47:00 <armax> and I’ll get it sorted 17:47:35 <Sukhdev> armax: OK - thanks - and, if all checks out OK, please approve it for merge 17:47:42 <armax> will do 17:48:06 <Sukhdev> armax: we have couple of back port patches for killo that require core approval - 17:48:33 <Sukhdev> armax: I am assuming this is from the infra core people, right? 17:48:55 <Sukhdev> armax: shall I go bug them on infra channel - or my understanding is wrong? 17:48:56 <armax> Sukhdev: no the stable team 17:49:02 <armax> so kyle, et al 17:49:05 <Sukhdev> armax: who is that? 17:49:12 <Sukhdev> kyle already approved it? 17:49:23 <armax> ihar or gary would be the others 17:49:23 <Sukhdev> so, it is you and doug then 17:49:59 <Sukhdev> armax: OK - I will make noise on neutron channel then - 17:50:27 <Sukhdev> armax: next one this - https://review.openstack.org/#/c/221932/ 17:50:51 <Sukhdev> do we want to approve it? I think this has to do with python 3, right? 17:51:11 <Sukhdev> I am not too sure about it 17:52:13 <Sukhdev> vikas: BTW, I have asked ash win to test this - https://review.openstack.org/#/c/222841/ 17:52:34 <Sukhdev> vikas: will let you know as soon we are done with the testing - or will approve it 17:52:40 <vikas> ok fine 17:53:09 <Sukhdev> armax: any thoughts on that patch? python 3 scares me - :-) 17:53:23 <Sukhdev> armax: or I am reading too much into this? 17:53:43 <armax> not sure I have teh full context 17:55:22 <Sukhdev> armax: this patch came from some infra - even I do not have context - 17:55:41 <Sukhdev> armax: since you are more plugged in, I thought you would know :-) 17:56:16 <armax> which one? 17:56:25 <armax> the bot patch? 17:56:27 <armax> that’s fine 17:56:31 <Sukhdev> yes 17:56:55 <Sukhdev> what is that for? 17:56:56 <armax> that’s a version bump that kyle posted a few days ago 17:57:06 <armax> he released a new version of the client 17:57:38 <Sukhdev> armax: Oh - I thought this has to do with python 3.0 support - 17:57:47 <armax> nop 17:57:50 <Sukhdev> armax: oh no worries then 17:58:02 <Sukhdev> armax: shall we approve it for merge then? 17:58:15 <armax> go ahead 17:58:33 <Sukhdev> armax: cool - 17:58:45 <Sukhdev> #topic: Mitaka discussion 17:59:30 <Sukhdev> armax: we do not have full quorum here - but, any thoughts about Mitaka for L2 GW? 17:59:59 <armax> not sure what’s the point of talking about mitaka 18:00:04 <armax> when nothing got done in Liberty 18:00:30 <Sukhdev> armax: I mean the charter for L2GW for mitaka 18:01:07 <Sukhdev> armax: i.e. do we want to plan any new features, use cases for L2GW for mitaka? 18:01:36 <gsagie_> :) 18:01:47 <armax> I think we lack the basics 18:01:57 <armax> and we didn’t tackle those in liberty 18:02:11 <armax> so really Mitaka should be a Liberty redux 18:02:25 <Sukhdev> armax: we got couple of critical enhancements done in Liberty 18:02:45 <armax> meh, that shouldn’t have taken the entire cycle, come on! 18:03:30 <Sukhdev> armax: understood - but, that also depends upon the resource availablity 18:03:37 <armax> we’re over time 18:03:47 <armax> agreed 18:03:52 <Sukhdev> armax: thanks for pointing it out - 18:04:01 <armax> so no resource availability no point on discussing mitaka thoughts 18:04:03 <Sukhdev> we are done - need to free up the channel 18:04:09 <Sukhdev> thanks folks for attending 18:04:17 <Sukhdev> bye 18:04:22 <Sukhdev> #endmeeting