17:02:13 #startmeeting networking_l2gw 17:02:13 Meeting started Mon Oct 12 17:02:13 2015 UTC and is due to finish in 60 minutes. The chair is Sukhdev_. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:02:14 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:02:17 The meeting name has been set to 'networking_l2gw' 17:03:07 #topic: Agenda 17:03:13 #link: https://wiki.openstack.org/wiki/Meetings/L2Gateway#Meeting_October_12.2C_2015 17:03:21 armax: are you here? 17:03:30 yes 17:03:37 fighting a gate breakage though 17:03:52 I do not see anybody from India team here 17:04:43 * eshedg slaps gampel around a bit with a large fishbot 17:05:20 Ofer - are you here? I do not know your IRC handle 17:05:24 hi sukhdev, i would like to add an item to the agenda 17:05:36 eran gampel & myself are attending, instead of Ofer 17:05:57 eshedg: welcome 17:06:15 Waiting for India team to join us 17:06:21 tnx. sure. 17:07:05 #topic: Release planning 17:07:35 We are down to three patches that need to be merged before we can release L2GW package 17:08:13 armax: we need your blessing on the three patches so that I can approve them for merge so that we can release it 17:08:30 Sukhdev_: which are they? 17:08:49 armax: https://review.openstack.org/#/c/202495/ 17:09:03 https://review.openstack.org/#/c/222841/ 17:09:22 and, https://review.openstack.org/#/c/206638/ 17:09:32 Sukhdev_: you’re not going to get my blessing on 202495 17:09:41 Sukhdev_: I think it is wrong 17:10:04 and so is 222841 17:10:09 as for 206638 17:10:24 I see no follow up 17:10:27 on my -1 17:11:24 I wonder if maruti is on PTO 17:11:37 russell’s point on switching to ovslib must be addressed if we want the l2gw project to be other to other contributions 17:11:43 we can’t ignore people’s feedback 17:11:49 especially if it’s valid one 17:12:13 I saw his email last week stating that he has addressed your comments - perhaps it was on a different patch 17:13:04 I think his email was about 202495 - sorry about confusion 17:13:30 His email seems to suggest he addressed your concerns - 17:14:35 armax: can you please respond to maruti's email and give him your feedback as to what he needs to address the concerns - so, that we can get this merged? 17:15:17 Sukhdev_: why? he said ‘done’ and ‘done' 17:15:23 Sukhdev_: maybe he forgot to push? 17:15:49 Sukhdev_: besides, without CI passing, do you really want to merge? 17:16:30 armax: I am confused - I see +1 from jenkins 17:16:43 not jenkins the 3rd party ci 17:17:40 Oh I see - since maruti is not here, it is bit hard...let me follow up with his email and lets see if we can get to some resolution on this 17:18:14 other than that I do not have anything additional on this.. 17:18:24 we can move on to the next topic - 17:19:04 #topic: L2GW as cloud-interconnect 17:19:14 eshedg: you are up 17:19:26 yep. gampel and me are here. 17:19:32 We think that with adding additional optional API to the L2 GW and small modification to the HW vtep schema this is possible 17:19:46 eshedg: in the last meeting we had asked for some kind of writeup on this 17:19:58 to use the L2 Gw API to connect Openstack cloud 17:20:14 yes i uploaded slide https://drive.google.com/file/d/0B6xagq1_AACadE40WVBXS196b2M/view?pli=1 17:20:31 * Sukhdev_ looking 17:20:53 #link: https://drive.google.com/a/arista.com/file/d/0B6xagq1_AACadE40WVBXS196b2M/view?pli=1 17:21:01 armax: you may want to look at this 17:21:12 * armax looking 17:22:40 I think what concerns me the most about this use case is the fact that there are too many ‘caveats' 17:23:01 like what? 17:23:12 if I udnerstand it correctly you have two totatlly independent openstack deployments 17:23:14 correct? 17:23:23 yes 17:23:25 and you want to take 2 neutron logical networks 17:23:43 and connect the data plane 17:23:53 so that it looks like a single broadcast doamin 17:23:55 domain 17:23:57 is that so? 17:24:30 yes but it could be discovery driven or DB (SDN) driven 17:24:38 ok 17:25:13 so, it could be openstack on both side or hybrid model 17:25:33 we think that if we add to the Vtep HW to the locater a tunnel id 17:25:43 the problem I envision is that these logical networks must be crafted in a way that they can actually work together 17:25:53 like IP spaces, gateways, etc 17:26:00 to not mention routing 17:26:22 okay, the use case we have in mind is the "tricircle" 17:26:34 where we have an orchestration "over cloud" that handles it 17:27:00 ok, without something on ‘top’ that abstract everything 17:27:22 this feature alone is too error-prone 17:27:23 The cascading project that create a shared network resource 17:27:48 we need validation process 17:27:53 i agree 17:28:10 so I’d rather see that take shape first 17:28:34 I wouldn’t personally want to expose such a capability with some overarching solution that ensure usability 17:28:47 *without 17:29:30 we agree that this functionality without a "top" is cumbersome 17:29:50 we could think on a authentication of the two sides that agree to connect the two networks 17:29:52 but there still is merit in it, if some user wants to connect manually multiple clouds, without a top management 17:30:32 eshedg: without proper safeguards, I bet that could be really hard to achieve successfully 17:30:38 there are way to many ‘gotchas' 17:30:43 i agree that this is a far-fetch 17:30:49 eshedg: I see the use case - but, as armax pointed out - the routing and IP management needs to be addressed 17:30:51 however, we wanted to re-use as much as we could with tricircle 17:31:33 if we could extend l2gw APIs for the tricircle "top cloud" use case, it would be better 17:32:11 otherwise, we will need to create the "border gateway" within tricircle project 17:32:19 eshedg: I am not familiar with tricircle - any pointers? 17:32:35 gampel will send it 17:32:46 how to do handle the default gateway on the network’s subnet? 17:32:55 host routes etc? 17:33:29 good question. we differentiate between "l2" cross-cloud connectivity 17:33:33 and "l3" connectivity 17:33:34 design doc of https://docs.google.com/document/d/19BXf0RhkH8wEEymE2eHHqoDZ67gnzgvpr3atk4qwdGs/edit?usp=sharing 17:34:01 if it's "l2", then it's a single broadcast domain 17:34:07 #link: https://docs.google.com/document/d/19BXf0RhkH8wEEymE2eHHqoDZ67gnzgvpr3atk4qwdGs/edit#heading=h.5r6zgqbiehsh 17:34:21 if it's "l3" then the subnets are different on each physical site 17:35:10 I would have to read through and digest these docs 17:35:25 okay, excellent 17:35:39 me too - I must agree that this is an interesting use case 17:35:48 I will review these documents as well 17:36:21 eshedg gampel: are you going to be in Tokyo? 17:36:24 yes 17:36:27 Yes 17:36:43 Can we set up some time to discuss this in person? 17:36:46 I can see the use case, my only concern is the usability 17:36:47 we have a talk about the multi site 17:37:14 sure yes 17:37:46 we try to address the usability so it will be seamless to the user 17:37:47 i suggest we discuss this in-person in tokyo, pick the time & place 17:37:59 armax, myself and other L2GW team will be there - it may be prudent to do a quick whiteboard discussion in Tokyo 17:38:04 before we rely on some f2f time 17:38:12 let’s put this on the next meeting agenda 17:38:21 so that we make sure people are reviewed the docs etc 17:38:43 armax: unfortunately, our next meeting will be after summit 17:38:44 okay, that'll be excellent 17:38:48 oh... 17:39:05 Can we discuss this over the mailing list 17:39:17 armax: I will forward these documents via email to all the L2GW team so that they can review off-line 17:39:22 Sukhdev_: ok 17:39:32 Ok thx 17:39:36 thanks 17:40:03 gampel eshedg: I will follow up with our internal email to L2GW team so that everybody reviews these documents 17:40:29 okay, we will add some more data about how the packet flows side-to-side 17:40:30 then we can pick a time to get together during summit to go over it and discuss in person 17:40:33 on both l2 and l3 use case 17:40:55 please add us both to the mailing list 17:41:13 eshedg: How soon can you update these - so that I send the updated documents 17:41:36 we'll do it tomorrow on the AM (our time) 17:41:38 eshedg gampel : please send me your emails at sukhdev@arista.com 17:42:13 once you have updated version, shoot me an email - and I can follow up with the rest of the team 17:43:33 #action: Sukhdev to follow up with the L2GW team about the L2GW cloud-interconnect use case 17:43:50 anything else? 17:44:03 from us, no, thanks 17:44:58 armax: anything from your side? 17:45:06 Sukhdev_: nop 17:45:14 cool - I guess we are done 17:45:19 thank you 17:45:20 thanks folks 17:45:28 bye 17:45:33 bye 17:45:34 #endmeeting