15:01:50 <yamahata> #startmeeting neutron_northbound 15:01:50 <odl_meetbot> Meeting started Wed Nov 2 15:01:50 2016 UTC. The chair is yamahata. Information about MeetBot at http://ci.openstack.org/meetbot.html. 15:01:50 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:50 <odl_meetbot> The meeting name has been set to 'neutron_northbound' 15:01:54 <yamahata> #chair mkolesni 15:01:54 <odl_meetbot> Current chairs: mkolesni yamahata 15:02:01 <yamahata> #topic agenda bashing and roll call 15:02:06 <yamahata> #info yamahata 15:02:19 <yamahata> #link https://wiki.opendaylight.org/view/NeutronNorthbound:Meetings meeting page 15:02:27 <mkolesni> #info mkolesni 15:02:48 <yamahata> Today I'd like to discuss about neutron stadium mainly and meeting timeslot 15:03:00 <yamahata> as follow up of openstack summit 15:03:07 <yamahata> do we have any other topics? 15:03:19 <mkolesni> what about the stable branch & release tag 15:03:36 <yamahata> Yeah. 15:03:37 <mkolesni> newton is already out we need to release as well 15:03:45 <yamahata> right. 15:04:44 <yamahata> so first release, stable branch and then timeslot, finally neutron stadium. Does it make sense? 15:05:03 <mkolesni> sure 15:05:16 <yamahata> #topic Announcements 15:05:26 <yamahata> last week there was openstack summit. 15:05:35 <yamahata> The biggest issue for networking-odl is neutron stadium 15:05:51 <mkolesni> as i see it stable branch is actually required by stadium 15:05:56 <yamahata> #link https://review.openstack.org/#/c/383910/ Ocata: Assessment for networking-odl 15:06:22 <yamahata> ODL community gathered and discussed on what we can do to remain in neutron stadium 15:06:36 <yamahata> by Ocata-1, Nov 14. 15:06:52 <yamahata> that's all from me. any other announcement? 15:06:58 <mkolesni> ok 15:07:17 <yamahata> #topic stable release/tag 15:07:45 <yamahata> Given the current situation, I'm thinking to cut/tag stable/release at Ocata-1. 15:08:14 <yamahata> because we're heavily working on newton backlong to fullfill neutron stadium requirement. 15:08:21 <mkolesni> actually stadium work doesnt relate to the release were suposed to make 15:08:26 <yamahata> Does it make sense? 15:08:36 <mkolesni> i think we need to release asap 15:08:55 <yamahata> I see. We have already release notes. So any other patches to include? 15:09:13 <mkolesni> maybe just the journal would be nice but we can always backport it 15:09:22 <mkolesni> as far as im concerned 15:09:34 <mkolesni> we can backport anything we see fit later on 15:09:38 <yamahata> Do you have list of patches? 15:10:02 <mkolesni> https://review.openstack.org/#/c/374613/ 15:10:09 <yamahata> Okay, so I'd like to include rally fix and grenade patches. 15:10:16 <mkolesni> #link https://review.openstack.org/#/c/374613/ Journal recovery for syncing with ODL 15:10:38 <mkolesni> ok so lets make an effort to push the last bits and get the release out 15:10:43 <yamahata> #link https://review.openstack.org/#/c/353050/ 15:10:44 <yamahata> Add grenade plugin 15:10:48 <mkolesni> im guessing version would be 3.0.0? 15:10:55 <yamahata> #link https://review.openstack.org/#/c/392005/ rally fix 15:11:02 <yamahata> Yes. 15:11:26 <mkolesni> ok 15:11:49 <yamahata> So after merging those three patches, let's ask to cut stable/newton 15:11:59 <yamahata> I'll upload the patch for that 15:12:36 <yamahata> #action yamahata after merging three patches, upload a patch to cut stable/newton 15:13:01 <mkolesni> cool 15:13:07 <yamahata> stable/branch is done. 15:13:10 <yamahata> next topic 15:13:15 <yamahata> #topic meeting timeslot 15:13:39 <yamahata> At openstack summit John asked me to consider the timeslot of irc meeting. 15:13:44 <yamahata> So I sent out a mail on it. 15:13:46 <mkolesni> seems that jjoyce is fine with Wed 15:14:10 <yamahata> #link https://lists.opendaylight.org/pipermail/neutron-dev/2016-November/001044.html 15:14:45 <mkolesni> however if not then Mon at the same time is fine for him 15:15:14 <yamahata> Oh Now I read the reply. 15:15:41 <yamahata> oh, either timeslot works for me too. 15:15:55 <yamahata> Monday or wednesday. 15:16:35 <yamahata> I'd prefer not to churn unnecessarily. If possible, I'd like to stay same time slot. 15:16:47 <yamahata> thoughts? 15:16:47 <mkolesni> sure if jjoyce is fine with it 15:17:38 <yamahata> I see. first option is wednesday, the second choice is monday. 15:17:53 <mkolesni> the only thing making monday better is that we have more time for AIs 15:17:54 <yamahata> #action yamahata reply john's mail. 15:18:09 <mkolesni> but i dont mind to keep it on Wed 15:18:18 <yamahata> Oh I see. Can you please reply to his mail for Monday? 15:18:33 <yamahata> Usually Monday is good for week planning. 15:18:40 <mkolesni> ok 15:19:16 <yamahata> My plan is to keep this time slot next week. and then switch timeslot if we'll do. 15:19:29 <mkolesni> sure 15:19:49 <yamahata> next topic 15:19:53 <yamahata> #topic neutron stadium 15:20:21 <mkolesni> ive been focusing on docs there 15:20:29 <yamahata> #link https://review.openstack.org/#/c/383910/ ocata assesment 15:20:36 <mkolesni> we also saw neutron-lib has CI and also it is being used 15:20:49 <mkolesni> do you know what else is needed on that front? 15:21:12 <yamahata> For now, nothing to do until Nov 14. 15:21:31 <yamahata> In middle/long term, we need to get involved to help neutorn-lib effort. 15:22:01 <mkolesni> do you know more details? 15:22:14 <mkolesni> i.e. are we supposed to not import anything from neutron? 15:22:22 <yamahata> Right. 15:22:37 <yamahata> So neutron-lib needs to be enhanced to cover more. 15:23:02 <yamahata> subprojects are expected to contribute to neutron-lib to make it happen. 15:23:03 <mkolesni> im not sure 100% is ever possible 15:23:21 <mkolesni> but we can try 15:23:45 <yamahata> neutron-lib is still at early phase. there are many thing to do. 15:24:24 <yamahata> any other questions? 15:24:33 <mkolesni> i think there maybe the decision could be changed since we do use it to some degree 15:24:56 <yamahata> Agreed. I expect so. 15:25:10 <yamahata> #link https://etherpad.openstack.org/p/Keeping_ODL_in_OpenStack_Stadium meeting minutes at openstack summit 15:25:24 <yamahata> #link https://bugs.launchpad.net/networking-odl/+bugs?field.tag=neutron-stadium-evolution task list for neutron stadium 15:25:40 <yamahata> #link https://docs.google.com/spreadsheets/d/1j9edHIQpLUA7NS9L8GugIQ8X38-LoZ5cvqsAEHdsz_0/edit#gid=0 details task/owner tracker 15:25:59 <mkolesni> ill keep working on the docs front 15:26:15 <mkolesni> still need to document full sync 15:26:24 <yamahata> Cool. 15:26:29 <mkolesni> and also have so deployer docs 15:26:37 <mkolesni> some* 15:26:51 <mkolesni> or admin docs 15:27:14 <yamahata> I suppose admin/user docs has higher priority than developer doc 15:27:32 <mkolesni> sure 15:28:17 <yamahata> for scenario tests, networking-odl has tested only v1 driver + old netvirt. 15:28:21 <mkolesni> ill get on top but my knowledge might be lacking in some areas so people should feel free to join 15:29:05 <mkolesni> re the docs 15:29:58 <yamahata> regarding to stable branch maintenance, there are three patches. 15:30:38 <yamahata> #link https://review.openstack.org/#/c/392286/ 15:30:38 <yamahata> tools: catch up neutron-lbaas change 15:30:50 <yamahata> #link https://review.openstack.org/#/c/389856/ stop doing any magic cloning of neutron during CI 15:31:01 <yamahata> #link https://review.openstack.org/#/c/389859/ stop doing any magic cloning of neutron during CI 15:31:27 <mkolesni> ill take a look though im not farmiliar with OS CI so much 15:31:30 <yamahata> are you interested in it? Ihar has complained about the current situation of networking-odl. 15:31:58 <yamahata> I see. Then, I'll continue to take care of those three. 15:32:13 <mkolesni> ok 15:32:24 <yamahata> I just wondered about role of Ihar and you. 15:32:45 <mkolesni> Ihar is focused on Neutron 15:33:01 <mkolesni> and is a stable maintainer 15:33:12 <mkolesni> so probably thats his interest 15:33:26 <yamahata> I see. 15:34:27 <yamahata> regarding to other tasks, the first priority is to make (all) scenario tests pass. 15:34:42 <yamahata> the second is to run multi-node tests (and hopefully make them pass) 15:34:52 <yamahata> the thirdone is to run functional tests. 15:35:01 <mkolesni> is there some ci infra problem there? 15:35:07 <yamahata> this one would be easy because it has only one test at the moment. 15:35:37 <yamahata> I don't see big issue. the task is to write a patch to project-config to define jenkins job. 15:35:40 * sridharg has a question regarding ovsdb_interface and would like to bring it up during the appropriate slot. 15:35:59 <mkolesni> hi sridharg :) 15:36:05 <yamahata> sridharg: ok. After this topic, we'll have open mike slot 15:36:14 <sridharg> sure thanks yamahata mkolesni 15:36:44 <yamahata> the last task is fullstack tests. I guess this would be difficult because we need to write test scheme from the scratch with defining what to test. 15:36:48 <mkolesni> so on Ocata M1 armax will redo his assesment? 15:37:20 <mkolesni> not sure what fullstack since were agentless 15:37:23 <yamahata> I suppose at Ocata-1, the patch, https://review.openstack.org/#/c/383910/ will be merged. 15:37:38 <yamahata> Until that we can keep give comment to update the situation. 15:37:59 <mkolesni> so what exactly full stack means for us? 15:38:49 <yamahata> #link http://docs.openstack.org/developer/neutron/devref/development.environment.html#testing-neutron 15:39:15 <yamahata> #link http://docs.openstack.org/developer/neutron/devref/development.environment.html#fullstack-tests full stack stests 15:39:59 <yamahata> In my understanding, in ODL case, deploy ODL and node with OVS. create netns to emulate vm. 15:40:15 <yamahata> send request to ODL neutron northbound etc. check if VM connectivity etc. 15:40:19 <yamahata> That's my understanding. 15:40:39 <mkolesni> ok this has little chance of happening in 2 weeks 15:40:41 <yamahata> It would be arguable that it's neutron test or ODL test though. 15:41:12 <yamahata> For functional test, we have only db migration sync validation tests at the moment. 15:41:34 <yamahata> In future, deploy neutorn server with networking-odl and deploy ODL with only neutron northbound. 15:41:40 <mkolesni> yea need more 15:42:03 <yamahata> send request to neutron server and check if ODL md-sal has expected data. 15:42:13 <yamahata> But it's after ocata-1. 15:42:21 <mkolesni> sure 15:42:44 <yamahata> any other issue to discuss? 15:43:17 <mkolesni> please review docs patches i sent 15:43:24 <yamahata> sure, off course! 15:43:29 <mkolesni> i think theyre good 2 go 15:43:51 <yamahata> let's move on 15:43:57 <yamahata> #topic open mike 15:44:03 <yamahata> sridharg: you're on stage 15:44:30 <sridharg> yamahata, ack 15:44:48 <sridharg> yamahata, I wanted to bring up an issue that I have seem with OpenStack Newton and master 15:45:10 <sridharg> basically in Newton release ovsdb_interface is defaulted to "native" (from "vsctl") 15:45:17 <yamahata> Yes. 15:45:23 <sridharg> mitaka - https://github.com/openstack/neutron/blob/stable/mitaka/neutron/agent/ovsdb/api.py#L33 15:45:28 <sridharg> newton - https://github.com/openstack/neutron/blob/stable/newton/neutron/agent/ovsdb/api.py#L33 15:45:50 <sridharg> My ODL setup is something like this - I have an external ODL Controller and an all-in-one devstack setup 15:46:25 <sridharg> when I'm creating some networks, dhcp agent is overriding the manager of ovs to local ip (i.e, ovsdb_connection) 15:46:42 <yamahata> Yeah, that causes issue. 15:46:44 <sridharg> this is causing issues in ODL 15:46:59 <yamahata> Have you looked at those patches, bug report? 15:47:11 <yamahata> #link https://bugs.launchpad.net/neutron/+bug/1614766 ovsdb managers shouldn't be erased/overwritten 15:47:22 <sridharg> I just noticed that you have a related patch - https://review.openstack.org/#/c/357528/2 15:47:25 <yamahata> #link https://review.openstack.org/#/c/357528/ ovsdb: don't erase existing ovsdb managers 15:47:40 <yamahata> #link https://review.openstack.org/#/c/382629/ bug/1614766 work around 15:47:56 <yamahata> For networking-odl, the work around is to use vsctl instead of native 15:47:58 <sridharg> yamahata, If I understand correctly the patch is trying to add a new manager instead of overriding the existing manager. Is this understanding correct? 15:48:09 <yamahata> That's right. 15:48:29 <yamahata> But in all-in-one node case, 6644 port is conflicting. So we need to change its port. 15:48:44 <sridharg> I agree with you that using "vsctl" would be a solution. 15:49:25 <sridharg> does having multiple managers configured on OVS work? 15:49:33 <sridharg> even when ODL Controller is an external controller? 15:50:02 <yamahata> Yes and no. 15:50:21 <yamahata> with external ODL case, we don't have port conflict. 15:51:18 <sridharg> ok. Say we have external ODL and dhcp-agent instead of overriding the manager adds a new manager (i.e., ptcp:127.0.0.1:6640) 15:51:45 <sridharg> how will it be able to create dhcp ports since ovsdb-server may not be listening on tcp? 15:51:56 <mkolesni> sorry guys gtg 15:52:10 <mkolesni> have a good day/evening 15:52:14 <sridharg> I've seen that in my setup, ovsdb-server is listening on unix socket and not tcp. 15:52:23 <yamahata> mkolesni: bye 15:52:27 <sridharg> i.e., --remote=punix:/var/run/openvswitch/db.sock 15:52:28 <mkolesni> bye 15:52:36 <yamahata> dhcp-agent will actively connect to 6640 port. 15:52:55 <yamahata> and directly talks ovsdb protocol. 15:53:03 <sridharg> I agree, but we should also have ovsdb-server listening on 6640 right? 15:53:12 <yamahata> Right. 15:53:28 <sridharg> By default (atleast in my setup) this is not the case. My ovsdb server is only listening on unix domain socket. 15:53:38 <yamahata> usually devstack sets it up. or other deployment tool like fuel should do. 15:54:14 <yamahata> With the current code, dhcp(or other) agent sets it up if ovsdb server doesn't. 15:54:23 <yamahata> and overwrites the existing setting. 15:55:56 <yamahata> #link https://github.com/openstack/neutron/blob/stable/newton/neutron/agent/ovsdb/native/connection.py#L80 15:56:39 <yamahata> #link https://github.com/openstack/neutron/blob/stable/newton/neutron/agent/ovsdb/native/helpers.py#L29 15:57:03 <sridharg> If i understood correctly, enable_connection_uri is only setting the manager. Does it also restart ovsdb-server to listen on tcp:6640? 15:57:22 <yamahata> we don't have to restart ovsdb-server. 15:57:36 <yamahata> Once manager is set, ovsdb will do it without restarting. 15:58:45 <sridharg> ooh ok. So your patch is going to help when we have an external ODL controller (for local ODL Controller we will have port-conflict). Is this correct? 15:59:04 <yamahata> that's right. 15:59:19 <yamahata> the patch is not high priority for me because we have work around. 15:59:29 <yamahata> If you like, please go take it over. 16:00:00 <sridharg> cool. thanks yamahata for patiently answering the queries. 16:00:12 <yamahata> any other issues? 16:00:18 <yamahata> or can we close the meeting? 16:00:29 <sridharg> yamahata, I will take a look at the review comments in your patch and will update it if required. thank you. 16:00:40 <sridharg> yamahata, ack from my side. 16:00:46 <yamahata> thanks 16:00:51 <yamahata> let's close the meeting. 16:00:58 <yamahata> #topic cookies 16:01:02 <yamahata> #endmeeting