17:02:08 #startmeeting networking_l2gw 17:02:08 Meeting started Mon Nov 23 17:02:08 2015 UTC and is due to finish in 60 minutes. The chair is Sukhdev_. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:02:10 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 17:02:12 The meeting name has been set to 'networking_l2gw' 17:02:23 oferby : hi 17:02:33 lets give some time for others to join in 17:02:38 sure 17:02:53 armax : are you here? 17:03:26 #topic: Agenda 17:03:33 #link: https://wiki.openstack.org/wiki/Meetings/L2Gateway#Meeting_November_23.2C_2015 17:04:44 anybody here from india team? 17:05:20 oferby: sorry for the confusion last week 17:05:35 this meeting takes place every other week 17:05:40 its OK. not a problem. 17:06:37 according to the schedule, it should have been last week. 17:07:21 Oh really - probably some kind of screw up then - sorry about that 17:07:36 I was looking for armax, I can not find him 17:08:00 he is not responding - I was going to discuss with him the release plan for Liberty 17:08:21 I have logged a bug to kick off this some time today or during this week 17:08:40 #link: https://bugs.launchpad.net/networking-l2gw/+bug/1518794 17:08:40 Launchpad bug 1518794 in networking-l2gw "Release request for networking-L2GW for stable/liberty" [Medium,Confirmed] - Assigned to Sukhdev Kapur (sukhdev-8) 17:08:52 until all will join, I understand that in the meeting you had with Ayal in Tokyo you / armax requested IDL. I started to work on it. 17:09:27 oferby : since armax is not here - lets get to the topic on the agenda to cover your use case 17:09:46 sure. 17:09:52 the change it actually easy 17:09:54 #topic: L2GW as inter-cloud interconnect 17:10:20 oferby : yes I met with the team in Tokyo 17:10:36 we need to add tunnel_key to Physical_Locator so we will be able to use single dst IP with multiple VNI 17:11:54 the hardware_vtep spec state that it is optionaly in the future (at the time it was written) 17:12:49 oferby : correct - we discussed about it in Tokyo 17:13:21 oferby : BTW, are you familiar with this bug -https://bugs.launchpad.net/networking-l2gw/+bug/1489095 17:13:21 Launchpad bug 1489095 in networking-l2gw "Switch to ovslib" [Undecided,New] 17:13:21 #link: https://bugs.launchpad.net/networking-l2gw/+bug/1489095 17:14:36 we can assign it to me. I already started to work on it 17:15:20 oferby : cool - thanks. 17:15:31 sure. 17:16:08 so, what armax and I discussed with the team was that we should start with a good foundation - and part of it was to switch to this library 17:17:17 this is what I was told, so I already started and have the lib, connection and model ready. I'm not working on commands and than I will change the code to use the new lib. 17:17:44 I'm now .... 17:18:05 oferby: So, what I was thinking is that we can release the liberty with its present implementation - then we can switch to the library and build upon it for Mitaka release 17:18:35 oferby : cool - so, you are ahead of me in that case :-) 17:18:36 seems good idea to me too 17:19:46 :-) I think I will be ready with the IDL right after the liberty release ... 17:19:52 oferby : I think you are on the right path - keep working off the master branch. Once we cut the liberty release, then we can work on your patches 17:20:04 sure 17:20:31 maruti is not here - I wish he was. He had pointed out one issue with the library - 17:20:53 give me a sec - let me look as to the issue he raised 17:21:07 if there is something you would like to me look at, let me know. 17:22:37 maruti states that python OVS library does not have support for manager table yet 17:23:00 and we use manager table in our present implementation 17:23:51 oferby : are you aware of this? 17:23:54 I do not use the data model from the lib. I created a model for L2GW and I use the IDL for connection management and row manipulation 17:24:56 for the connection management - who initiates the connection? server or the agent? 17:25:11 agent 17:25:31 oferby : ah ha - that is the issue then - we need it both ways 17:25:43 this is for the writer. for the monitoring - I did not looked at yet. 17:26:03 we will have something similar to the writer/monitor 17:26:20 present implementation supports both ways - through config 17:27:02 I would like the connection to be configured using api command instead of config file 17:27:03 anyways, he brought this issue to armax and my attention 17:27:35 for a generic implementation, we need to support both ways - 17:27:53 in inter-cloud it is very important to be able to easily add gw 17:28:30 so, when you add a gw, what are the steps you will take? 17:29:28 state its connection string and it will connect. something like tcp:192.168.100.1:6640 17:29:50 an agent will be able to manage multiple connections to ovsdb servers 17:30:05 oferby : where will you set this connection string ? 17:30:14 api command 17:30:56 so, does this API write it into the OVSDB server DB? 17:31:39 no. to the neutron db for persistence. the agent will read it when starting 17:32:06 got it 17:32:33 there was a patch recently pushed to support connection to multiple OVSDB servers - are you aware of it? 17:33:25 no, but we are going to change to the OVS lib anyway so the connection will be different anyway 17:33:44 I just approved it yesterday :-) 17:33:53 here it is just FYI - https://review.openstack.org/#/c/222841/ 17:34:00 :-) 17:34:26 so, coming back to the issue that I was describing - 17:35:10 there is a requirement (and we recently added support for it) to support the initiation of the connection from the server as well 17:35:43 most of the controllers work in that fashion - 17:35:44 I must say that I'm not sure if there is support in the IDL for that. I will check. 17:36:20 i.e. you configure the IP of the agent in the OVSDB server and it will automatically connect 17:36:33 this makes deployment very easy for the operators 17:36:56 And, for that we use management table 17:37:08 Yes, please have a look 17:37:13 I see. I will take a look 17:37:41 In order to keep things backward compatible, we need to ensure that present functionality is not broken 17:37:51 sure 17:38:25 other than that I do not have anything to share 17:38:47 unless you have anything else to discuss, we can call it off now. 17:39:06 I will wait with inter-cloud after I will finish the IDL. ok? 17:39:11 I will follow up with armax off-line and get the release logistics worked out and will post on the ML 17:39:26 oferby : sounds good 17:39:46 ok. so we can call it a off. 17:39:55 oferby : BTW, can I ask you for a favor 17:40:01 sure 17:40:33 there were couple of link to the inter-cloud that your folks provided in Tokyo - 17:41:02 I can't seem to find them ( I probably deleted it accidently) - can you share those? 17:41:24 I would like to put them on the L2GW wiki for the benefit of others 17:41:37 sure. I will do that tomorrow my morning. 17:41:49 cool - thanks 17:42:11 Thanks for joining - have a wonderful day 17:42:13 bye 17:42:20 you too. bye. 17:42:25 #endmeeting