16:04:47 #startmeeting networking_ml2 16:04:48 Meeting started Wed Jul 15 16:04:47 2015 UTC and is due to finish in 60 minutes. The chair is rkukura. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:04:49 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:04:52 The meeting name has been set to 'networking_ml2' 16:05:01 #topic Agenda 16:05:25 #link https://wiki.openstack.org/wiki/Meetings/ML2#Meeting_July_15.2C_2015 16:05:43 anything to add to today’s agenda? 16:06:10 #topic Announcements 16:06:38 Liberty-2 is in about two weeks 16:06:50 #link https://wiki.openstack.org/wiki/Liberty_Release_Schedule 16:07:00 Any other announcements? 16:07:24 #topic Action Items 16:07:53 Only action item was for me regarding planning the mid/end/pre -cycle sprint 16:08:19 I’ve updated the etherpad with possible dates in September and early October 16:08:31 #link https://etherpad.openstack.org/p/Neutron_ML2_Mid-Cycle_Sprint 16:08:38 rkukura: did we select the venue - we had three choices 16:09:19 Sukhdev: We are targetting the San Jose area, since most people whove expressed interest are local there 16:09:37 But finding a date seems much more difficult than finding a host in that area 16:10:05 I mean we have choice of hosting at Cisco, Yahoo, or Brocade 16:10:06 I request that anyone interested in participating add their availability to the last section of the etherpaf 16:10:09 etherpad 16:10:36 Sukhdev: right, plenty of good choices, but I’m not sure anyone can commit to hosting without knowing the dates 16:11:09 makes sense 16:11:27 so, lets finalize the dates first 16:11:45 I’ll followup on the existing email thread to encourage people to update the etherpad with their availability ASAP 16:12:18 Based on my constraints, best weeks are early October, but there are possibilities in September 16:12:34 Anything else on this or other action items? 16:13:08 I’d like to commit to a date by next week’s IRC meeting and then finalize who is hosting. 16:13:33 #topic L2 Modular Agent discussion 16:13:56 I don’t see banix, who has been driving this. Does anyone else have an update? 16:14:53 We can get back to this if banix shows up 16:15:04 #topic Physical Topology Discussion 16:15:45 asomya: Anything we need to cover on this today? 16:16:19 guess not ;) 16:16:35 looks like folks are missing today :-) 16:16:51 asomya: You are back! 16:17:41 Anything on topology, or should we move on? Seems neither shivharis nor asomya is available. 16:18:15 we can get try this one later 16:18:34 #topic Ironic/Neutron-ML2 integration discussion 16:18:41 fyi - asomya having issues w/ irc ... he's still trying 16:18:54 rcurran: Thanks - we can get back to topology later 16:19:10 Sukhdev: Any update on ironic? 16:19:17 rkukura: Yes 16:19:50 Last time we ran out of meeting so, I thought I put this back so that folks can have questions 16:20:21 basically, we have finalized most of the interface related issues 16:21:01 Since this is an interface to ML2, I thought it is appropriate to discuss it here 16:21:30 The address to the wiki is - https://wiki.openstack.org/wiki/Meetings/Ironic-neutron 16:21:45 there are two specs - one is approved, other one is ready to go 16:22:13 I have one question for this team - that has to do with the test framework 16:22:45 Do we have any specific test framework in place which specifically tests the ML2 interface? 16:23:19 Sukhdev: Do you mean the driver APIs? 16:23:21 what does "ml2 interface" mean? driver interface? 16:23:41 yes - the driver interface 16:24:16 for instance, to test/verify the validity of specific contents of ML2 structures 16:24:19 Sukhdev: There are tests that use a test driver with a bunch of asserts that validate that the driver APIs are called properly, but these are not ideal 16:25:42 I am thinking about scenario test for Ironic interface - but, thought if something is already in place, perhaps we can leverage it 16:27:16 for scenario test, I am thinking to create ovs bridge and launch a vm in a same host. I am not sure how ML2 driver interface test works for this. 16:28:20 amotoki: not sure what is available that can be leveraged - hence, I asked the question 16:29:02 The intent will be to setup things from the Ironic side, and launch and instance and verify that all the ML2 related stuff is correctly received and then verify 16:29:50 the network switch logic - i.e. DHCP is correctly assigning the address for PXE network and when the network flips, the correct tenant IP address is assigned 16:30:06 Sukhdev: I see. I think it is a bit differnet from testing a ml2 driver interface. 16:30:09 something of that nature is ideally need to test this functionality end-to-end 16:30:34 Sukhdev: Since these tests span various projects, I don’t think the existing driver API functional tests are relevant. 16:31:05 amotoki: there are two parts to it - one part is verify that the Ironc-neutron interface is correct - that is where we need to verify the ML2 content and second part is network flip logic 16:31:40 for the second point, we can test how a mechanism driver processes port profile and others. 16:32:12 sorry, it is not related to the second topic.... 16:32:53 amotoki: I am trying to understand if we have anything in place today which can be leveraged here and built upon? 16:33:42 Sukhdev: I think we don't have any functional level testing framework to verify if ML2 driver works as expected. 16:34:27 Sukhdev: Seems you’ll need a test-specific mechanism driver, somewhat like neutron.tests.plugins.ml2.driver.mechanism_test.TestMechanismDriver, but for ironic. 16:34:39 I was afraid that was the answer - wanted to double check in case I was not looking at the right place :-) 16:35:42 rkukura: thanks for pointing this out - this could be the starting point- let me check into this 16:36:14 I think what we would like to test is how a switch port (e.g. ovs port) is handled in ML2 driver. right? 16:36:25 Sukhdev: Unfortunately, this test driver uses regular python asserts rather than the test framework’s asserts. 16:36:35 amotoki: correct - you got it 16:36:59 rkukura: oh I see 16:37:08 Anything more on ironic? 16:37:34 rkukura: no thanks - I am done 16:37:59 Any questions for Sukhdev on this? 16:38:04 please go ahead 16:38:17 Ok, we’ll move on 16:38:26 #topic Task Flow Document 16:38:39 #link https://docs.google.com/document/d/1aSgTVB7nW_v7lHH0Z0DUgfymEsx0O16k1Jgu7QFXkFA/edit?usp=sharing 16:39:23 I just want to mention that some of Cisco’s driver maintainers are starting to look closely at this, and hopefully we’ll get some useful feedback 16:40:37 I suggest everyone interested in the sync/async work planned for the sprint continue to work on this in the meantime. I intend to. 16:40:48 Anything else on this? 16:41:12 asomya: Did you want to discuss Physical Topology? 16:41:55 rkukura: Not much changed since last week 16:42:18 #topic Open Discussion 16:42:18 asomya: how is your doc coming along 16:42:23 #undo 16:42:23 Removing item from minutes: 16:42:44 asomya: you should share - even if it is preliminary stage - 16:42:58 Sukhdev: We're adding to that doc as we go along, we'll definitely share once it's in a readable format :) 16:43:24 asomya: any ETA? 16:44:04 Sukhdev: Don't have a hard date right now unfortunately 16:44:27 asomya: Do we still want to include topology in a [mid/late/pre]-cycle meeting in September or early October? 16:44:44 asomya: no worries - if you need any help, we are here….even if you need to discuss verbally….. 16:44:57 rkukura: We can keep it as a discussion item for the late-cycle 16:45:28 Sukhdev: much appreciated, we will definitely want community feedback right after we have teh basic framework ready 16:45:35 asomya: OK. Please add your availability to https://etherpad.openstack.org/p/Neutron_ML2_Mid-Cycle_Sprint 16:45:42 Anything else on this? 16:45:53 #topic Open Discussion 16:46:03 rkukura: sure, will add myself 16:46:09 asomya: thanks 16:46:11 Anything else to discuss today? 16:46:33 If not, we’ll wrap up a bit early :) . 16:47:22 rkukura: thanks 16:47:30 armax: Did you want to bring up anything regarding the driver APIs and clarifying responsibility during methods like create_network_precommit? 16:48:17 rkukura: I was considering adding this to the agenda - but, armax mentioned he was not available - hence, did not add it 16:48:49 Sukhdev: OK, he can join us next week if he’d like to discuss it 16:48:54 Thanks everyone! 16:49:01 #endmeeting