16:03:05 <yamahata> #startmeeting neutron_northbound
16:03:05 <odl_meetbot> Meeting started Fri Jan 29 16:03:05 2016 UTC.  The chair is yamahata. Information about MeetBot at http://ci.openstack.org/meetbot.html.
16:03:05 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:03:05 <odl_meetbot> The meeting name has been set to 'neutron_northbound'
16:03:12 <yamahata> #topic agenda bashing and roll call
16:03:18 <yamahata> #info yamahata
16:03:37 <yamahata> #link https://wiki.opendaylight.org/view/NeutronNorthbound:Meetings#Agenda_for_Next_Meeting_.281.2F29.2C_2016.29 meeting agenda
16:03:55 <yamahata> is there any agenda to add?
16:05:26 <yamahata> seems nothing. move on.
16:05:28 <yamahata> #topic Announcements
16:06:01 <yamahata> opendaylight minisummit at OSN will be held
16:06:08 <yamahata> It's CFP is also open
16:06:24 <yamahata> #link https://lists.opendaylight.org/pipermail/discuss/2016-January/006134.html CFP for ONS OpenDaylight Mini Summit
16:06:41 <yamahata> opendaylight summit is also announced
16:06:55 <yamahata> #link http://events.linuxfoundation.org/events/opendaylight-summit opendaylight summit
16:07:39 <yamahata> lastly I can't chair next week. Feb 5. If anyone else is willing to chair, it would be skipped.
16:07:46 <yamahata> anything else to announce?
16:09:06 <yamahata> #topic action items from last meeting
16:09:33 <yamahata> For boron planning, I announced some links at odl neutron dev mailing list
16:09:45 <yamahata> #link https://wiki.opendaylight.org/view/NeutronNorthbound:FutureReleaseWishList
16:10:02 <yamahata> #link https://docs.google.com/spreadsheets/d/1b5JuGWwYISno-mcb92SC3_1bwncfo4ToB8b8480D15Q/edit?pref=2&pli=1#gid=0 feature and resource assignment
16:10:13 <yamahata> I created this google-sheet mimicking ovsdb planning.
16:10:52 <yamahata> But I'm not sure we need such formal assignment because our engineering resource is small and scope is smaller.
16:11:42 <yamahata> Maybe to track progress, trello board may be better. But I haven't decided yet.
16:12:06 <yamahata> john_a_joyce:  reviewed the patch.
16:12:16 <yalei> it is for odl or include neutorn?
16:12:23 <yamahata> #link https://review.openstack.org/#/c/215612/
16:12:23 <yamahata> Enable vhost-user ports on supported platforms.
16:12:34 <yamahata> it's networking-odl.
16:12:46 <yalei> yamahata: thanks
16:12:49 <yamahata> john_a_joyce: are you fine with the patch?
16:13:13 <yamahata> in order to make port-binding flexible, the follow up patch is coming
16:13:22 <yamahata> #link https://review.openstack.org/#/c/273390/
16:13:22 <yamahata> Configure valid VIF types.
16:13:24 <john_a_joyce> yes i reviewed that - i find it acceptable
16:13:37 <yamahata> If you're fine, I'm going to merge it.
16:13:47 <yamahata> Okay, then I'll merge it.
16:14:12 <yamahata> done.
16:14:26 <john_a_joyce> yes i am fine
16:15:36 <yamahata> For boron planning Isaku and vithar have started the discussion.
16:15:53 <yamahata> it will be held on here, wiki page(wishlist page) and mailing list.
16:16:33 <yamahata> okay, all action items are done.
16:16:39 <yamahata> #topic patch reviews
16:16:50 <yamahata> I had one patch review item and it's done right now.
16:17:11 <yamahata> any other review for attention?
16:18:00 <yamahata> #topic bug
16:18:13 <yamahata> any bug needing care?
16:19:16 <yamahata> okay move on
16:19:21 <yamahata> #topic ML2 ODL driver rewrite
16:19:33 <yamahata> any raises?
16:19:41 <yamahata> I suppose patch review is on-going
16:20:18 <rcurran> yes, thanks for the comments
16:20:55 <yamahata> rcurran: I have question. You create one more thread for sync. Are you seeing difference?
16:21:12 <yamahata> I understand we'd like to make different service isolated from each services.
16:21:25 <rcurran> difference in what way?
16:21:35 <yamahata> e.g. L3 from L2
16:21:42 <rcurran> we know we have to support multiple journal threads to support HA
16:21:55 <yamahata> Can you elaborate on it?
16:22:21 <rcurran> in HA there will be as many threads as there are neutron controller deployments
16:22:49 <yamahata> yes, many threads in neutron servers.
16:23:59 <yalei> yamahata: do you mean we can use the same thread ?
16:24:20 <yalei> yamahata: to deal with L2 and L3
16:24:48 <yamahata> yalei: Yes. But we can have thread pool to handle journal db table.
16:25:08 <yamahata> rcurran's patch specifically uses two threads.
16:25:51 <yamahata> So I'm wondering if he has special reason to have _two_ thread instead of thread pool.
16:26:13 <yamahata> Or single thread is okay.
16:26:24 <rcurran> as written now it uses one thread/neutron-server process. i can eliminate the L3 call to start a thread but the overall design must be able to support multiple journal threads
16:26:42 <rcurran> in HA we don't have much of a choice
16:27:36 <yamahata> rcurran: I see. what kind of problem do you want solve?
16:27:53 <yamahata> e.g. request piling up in journal?
16:27:57 <rcurran> the case where two or more journal threads are running
16:28:07 <yamahata> or better isolation resource. e.g. L3 from L2
16:28:29 <yalei> is there some special reason to isolation?
16:29:11 <rcurran> i don't think we should separate the l3/l2 processing. odl controller doesn't care. neutron separatation is why l3/l2 is processed differently today
16:29:25 <rcurran> ml2 core_plugin vs. l3 service_plugin
16:30:29 <yalei> does the two kind of threads do the same?
16:30:37 <rcurran> yalei, no
16:30:49 <yalei> I see
16:31:21 <rcurran> a journal thread doesn't distinguish that it's processing, just that the rows are processed in the correct order (dependency checking)
16:31:29 <rcurran> that -> what
16:32:02 <rcurran> i did think to do that at one point, but see the payoff
16:32:08 <yalei> yes, that's what I mean
16:32:13 <rcurran> but didn't see the payoff
16:32:31 <yamahata> did or didn't?
16:32:46 <rcurran> didn't - why seperate l2 and l3
16:33:14 <rcurran> let me try again :-) ... i did at one point think to seperate ... but didn't see any payoff
16:33:25 <yamahata> Got it.
16:33:36 <rcurran> from odl controller perspective .. it just cares about the neutron events
16:33:54 <yamahata> So for now you have two threads, but for future, you'd like to switch thread pool. Correct?
16:34:56 <yamahata> At least it should be investigated.
16:34:58 <yamahata> ?
16:35:12 <rcurran> regarding number of threads, asomya and i are still looking into some changes ... but a requirement is that we *have* to suport two or more journal threads to support HA
16:35:55 <rcurran> so currently having two is actually good since it allows us to test (without having an HA environment)
16:35:56 <yamahata> Okay. Can you please elaborate the behind reasoning for the requirement?
16:36:37 <yamahata> What led you to play with thread?
16:36:38 <rcurran> i don't understand. in HA are you seeing a design w/ only one thread?
16:37:26 <yalei> what's the difference between the thread pool and two explicit thread?
16:37:38 <yamahata> No. I think we need multiple thread at some point.
16:38:25 <yamahata> In L3 plugin code, can you please some comment to explain the situation?
16:38:54 <yamahata> Otherwise, I thought there is a strong reason to use _two_ thread.
16:39:24 <yamahata> Maybe TODO or something
16:39:44 <rcurran> TODO stating that the call from L3 is starting 2nd thread?
16:40:10 <yamahata> Yes. e.g. future direction is to investigate about the number of threads.
16:40:36 <rcurran> ok
16:41:26 <yamahata> any other review issues?
16:42:19 <yalei> does the v2 driver support the full sync?
16:42:39 <yamahata> yalei: I haven't started it yet.
16:42:54 <yalei> yamahata: I want to investigate it.
16:43:00 <yamahata> yalei: cool!
16:43:15 <yamahata> give it a shot!
16:43:32 <yalei> yamahata: :)
16:43:46 <yamahata> anything else?
16:44:27 <yamahata> #topic OpenStack release support
16:44:39 <yamahata> no update so far.
16:44:52 <yamahata> #topic Beryllium release preparation and Boron planning
16:45:11 <yamahata> I've started discussion with other projects.
16:45:17 <yamahata> We can continue.
16:45:28 <yamahata> #topic Open Mike
16:45:32 <yamahata> any topics?
16:46:15 <vthapar> I raised a bug earlier today that I'd like to discuss: https://bugs.opendaylight.org/show_bug.cgi?id=5137
16:47:02 <vthapar> it is related to 4810, fixing model definitions.
16:47:05 <yamahata> Oh, extraroutes.
16:47:46 <vthapar> I was testing to see if it is supported so could use in vpnservice or not.
16:48:11 <vthapar> routes are defined as type string when should be host-routes.
16:48:29 <yamahata> 4810 seems wrong number.
16:48:40 <yamahata> vthapar: do you need it for Beryllium?
16:49:32 <vthapar> yamahata: I think I do.
16:50:16 <vthapar> I was testing adding multiple nexthops with same prefix, which isn't needed in Be. but found that basic functionality isn't working
16:51:19 <vthapar> also, it showed that we need to strengthen our test cases. they seem to be just for coverage, we're not doing proper validation.
16:52:00 <yamahata> Do you mean test in openstack side? tempest? or ODL ci?
16:52:33 <vthapar> I meant our ITs.
16:52:55 <vthapar> unless you believe these are taken care of in ODL CI, or tempest.
16:54:03 <vthapar> we just test that crud succeeds, shouldn't we also test that all the data we created/udpated is reflected correctly in mdsal?
16:56:05 <yamahata> Right, the current test doesn't check its actual value.
16:56:17 <yamahata> This bug is a blocker for vpnservice?
16:56:48 <yamahata> do you have (WIP) patches? or need help?
16:56:53 <vthapar> not blocker, but impacts a use case we'd like to support.
16:57:04 <vthapar> I am working on patch and wanted some inputs on that.
16:57:21 <vthapar> we curently have NeutronSubnet_HostRoutes class which serves same purpose.
16:57:52 <vthapar> so what fix would be preferred, make a generic Neutron_HostRoutes class and use for routers and subnets, or create a new NeutronRouter_Routes class?
16:58:37 <vthapar> both have destination and next-hop fields.
16:58:38 <yamahata> The right direction is to support extraroute extension cleanly.
16:59:07 <yamahata> But for Beryllium, code s already freezed.
16:59:42 <yamahata> If NeutronSubnet_HostRoutes works for you (with minimul code change), it might be better.
17:00:18 <vthapar> yeah, that could work, will have to check the yang for l3 though, that might still need change.
17:00:46 <vthapar> will get back once I have a patch ready and can discuss this in gerrit.
17:00:52 <yamahata> Yeah, discussion based on actual patch will be more productive.
17:01:19 <yamahata> Don't hesitate to ping me
17:01:30 <vthapar> sure will.
17:01:47 <yamahata> any other topics?
17:01:49 <vthapar> and what about improving ITs? will take it up in Boron?
17:02:23 <yamahata> it's surely the item for Boron. We should raise it.
17:02:30 <vthapar> it would be a good low-hanging fruit for new folks.
17:02:51 <vthapar> add it to wishlist or covered by Sonar topics in there?
17:03:19 <yamahata> Please add it to wishlist to track it.
17:03:25 <vthapar> will do.
17:04:02 <yamahata> anything else?
17:04:14 * yamahata running over time...
17:04:25 <yamahata> going once
17:05:03 <yamahata> going twice
17:05:13 <yamahata> The next week the meeting will be skipped.
17:05:34 <yamahata> thank you everyone!
17:05:41 <yamahata> #topic cookies
17:05:42 <yalei> yamahata: thanks
17:05:46 <yamahata> #endmeeting