15:02:50 <yamahata> #startmeeting neutron_northbound 15:02:50 <odl_meetbot> Meeting started Mon Jul 10 15:02:50 2017 UTC. The chair is yamahata. Information about MeetBot at http://ci.openstack.org/meetbot.html. 15:02:50 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:02:50 <odl_meetbot> The meeting name has been set to 'neutron_northbound' 15:03:02 <yamahata> #chair mkolesni 15:03:02 <odl_meetbot> Current chairs: mkolesni yamahata 15:03:06 <yamahata> #chair rajivk 15:03:06 <odl_meetbot> Current chairs: mkolesni rajivk yamahata 15:03:16 <yamahata> #topic agenda bashing and roll call 15:03:19 <mkolesni> #info mkolesni 15:03:22 <yamahata> #info yamahata 15:03:31 <rajivk_> #info rajivk 15:03:47 <mkolesni> id like to talk about Pike-3 and PTG/OS summit 15:03:50 <yamahata> #link https://wiki.opendaylight.org/view/NeutronNorthbound:Meetings meeting 15:04:09 <mkolesni> and of course reviews 15:04:20 <yamahata> any other topics? 15:04:37 <rajivk_> yeah there are some patches 15:04:51 <rajivk_> Xiao want su to merge in ocata branch 15:05:13 <mkolesni> bug fixes? 15:05:17 <rajivk_> yes 15:05:35 <mkolesni> are they backported yet? 15:05:46 <rajivk_> no 15:06:07 <rajivk_> resource ordering and full sync and recover of all the resources 15:06:11 <mkolesni> he should report the bugs on ocata as well i believe 15:06:28 <mkolesni> if it has db migration script it wont be backportable 15:06:49 <rajivk_> no, they don't have db migration scripts 15:07:26 <mkolesni> ok then afaik he should report the bugs on ocata and backport the patches 15:07:41 <rajivk_> should i discuss about them now 15:07:54 <rajivk_> or we can discuss at the end of meeting? 15:07:57 <yamahata> let's discuss at patches/bugs topics 15:08:10 <rajivk_> ok 15:08:18 <yamahata> do we have other topics? can we move on? 15:08:41 <mkolesni> lets move on 15:08:43 <yamahata> #topic Announcements 15:09:00 <yamahata> openstack summit RFP deadline is approaching 15:09:25 <mkolesni> yamahata, are you going to the upcoming summit and PTG? 15:09:42 <yamahata> For now, it's not decided unfortunately. 15:09:50 <mkolesni> even ptg? 15:09:51 <yamahata> It would take a while to decide. 15:09:54 <yamahata> Right. 15:09:59 <mkolesni> ok 15:10:07 <mkolesni> i will probably be going to ptg 15:11:22 <mkolesni> i know rajivk_ wants to go to summit 15:11:28 <mkolesni> at least send in papers 15:11:36 <rajivk_> I don't know yet. 15:11:55 <rajivk_> yes, i am planning to present something. 15:11:56 <yamahata> the deadline is less than a week. We need to be in hurry 15:12:01 <mkolesni> btw does ptg have some cfp equivalent? 15:12:14 <yamahata> No. ptg doesn't have presentation. 15:12:26 <mkolesni> but probably some planning? 15:12:36 <mkolesni> etherpad? 15:12:40 <yamahata> Right. 15:12:49 <mkolesni> do you know the details? 15:12:52 <yamahata> There would be discussion on the mailing list. 15:13:03 <mkolesni> ok 15:13:04 <yamahata> RIght now I don't see discussion started yet. 15:13:28 <mkolesni> yamahata, did you submit anything to summit? 15:13:41 <yamahata> I'm planning one proposal. 15:13:55 <yamahata> We can propose another jointly. 15:13:59 <mkolesni> what would you be proposing? 15:14:17 <mkolesni> we should sync up to avoid proposing the same things twice 15:14:32 <yamahata> about scalability evaluation. 15:14:35 <mkolesni> im not sure if i will go to summit but i might propose something 15:14:40 <yamahata> For details, we can discuss it offline. 15:14:53 <mkolesni> regarding ODL+OS or just OS? 15:14:56 <yamahata> #action yamahata sent out mail to discuss Openstack summit RFP 15:14:58 <yamahata> ODL + OS 15:15:04 <rajivk_> yamahata, can i join as well? 15:15:06 <yamahata> it's openstack summit. 15:15:18 <yamahata> rajivk_: off course. 15:16:06 <yamahata> any other announcement? 15:16:08 <mkolesni> ok if you have something in mind for proposal im open to suggestions 15:16:16 <mkolesni> can we talk about Pike-3? 15:16:19 <rajivk_> i want to propose something like, what are the parameter we should consider for openstack deployment 15:16:21 <yamahata> Pike-3 is Jul 24-28 as feature freeze 15:16:32 <yamahata> #topic Pike planning 15:16:49 <mkolesni> so we need to get all non-bugs done until then 15:17:09 <mkolesni> after that only bug fixes should be merged i.e. no more rfes 15:17:36 <mkolesni> we need to leave time for qe to test the driver 15:18:09 <yamahata> yeah. From Pike, networking-odl follows release-by-cycles 15:18:54 <mkolesni> ok so we have plenty of rfes still on the fly that we need to get in 15:19:23 <mkolesni> re rajivk_'s work on refactor its coming along nicely 15:19:38 <mkolesni> im concerned about https://review.openstack.org/#/q/status:open+project:openstack/networking-odl+branch:master+topic:bp/dep-validations-on-create 15:19:42 <yamahata> as big topic we have several v2driver staff, dhcp-port-creation, etc 15:20:27 <yamahata> Regarding to v2driver, we can drive optimistic lock patch and making the number of journal thread configurable 15:20:46 <yamahata> Regarding to pre-dependency calculation, we haven't reached consensus. 15:20:48 <mkolesni> i separated it from the branch as you requested 15:20:58 <mkolesni> https://review.openstack.org/467060 15:21:05 <mkolesni> i also found a bug when doing so 15:21:11 <mkolesni> https://review.openstack.org/482042 15:21:45 <mkolesni> the bug was due to an oversight on an earlier code refactor but it became very apperant in the optimistic locking patch 15:22:00 <mkolesni> so if you can review those 2 thats great 15:22:16 <yamahata> okay. 15:22:18 <mkolesni> i tested it locally with rally and there's no hinderance in performance 15:22:40 <yamahata> Regarding to 467060, rajis's review requests tests 15:23:00 <yamahata> The modification itself looks okay to me. 15:23:08 <mkolesni> yes i altered the test 15:23:15 <yamahata> Similar to 482042. 15:23:18 <mkolesni> there is no other way to test it besides using a mock 15:24:20 <yamahata> I see. then we can use mock. 15:24:25 <mkolesni> re 482042 there is no way to cause the bug to happen in a unit test, the only test that can be made is perhaps to prevent it in the future but if we change to dependency calculation it will become moot 15:24:31 <yamahata> anyway in detail we can continue discussion at gerrit. 15:24:44 <mkolesni> so i would rather not spend extra time on that if its not necessary 15:24:46 <mkolesni> sure 15:26:01 <yamahata> For Pike right now openstack CI is in very bad shape. 15:26:08 <yamahata> we need to address it for Neutron Stadium. 15:26:22 <yamahata> Especially tempest jobs are failing for long time. 15:26:39 <mkolesni> do we still need all the v1 tests? 15:26:39 <manjeets> timeouts in tempest job ? 15:26:55 <yamahata> ssh via floating ip is still failing. 15:27:11 <yamahata> timeout sometimes happens. but we can issue recheck. 15:27:25 <mkolesni> i think we should focus on fixing carbon v2 15:27:36 <mkolesni> im not sure we need to focus on v1 CI failure 15:27:38 <manjeets> from last 2 weeks the timeout failure frequency has increased 15:28:07 <yamahata> The reason for timeout is that testcases were enabled. 15:28:44 <yamahata> it added more test cases with ssh which added 5mins. 15:29:36 <yamahata> next one, networking-odl documentation needs to be reorganized. 15:29:42 <yamahata> to follow new standard 15:30:53 <yamahata> The topic name is doc-migration 15:31:04 <yamahata> #link https://review.openstack.org/#/q/topic:doc-migration 15:31:40 <yamahata> anything else for pike cycle? 15:31:44 <mkolesni> does this need to happen before Pike-3? 15:32:33 <yamahata> I suppose no. we can address it after Pike-3 15:32:59 <yamahata> and it's just to shuffle doc tree 15:32:59 <mkolesni> ok sounds reasonable 15:33:22 <mkolesni> btw do you know when the CI started failing? 15:33:52 <yamahata> very early in Pike 15:34:15 <yamahata> several bugs in ODL netvirt were fixed. but we're seeing failure still. 15:34:39 <mkolesni> i remember v2 carbon used to work fine 15:34:42 <mkolesni> and carbon isnt new 15:34:59 <manjeets> tempest smoke test for pinging an instance with fip is still failing 15:35:00 <yamahata> #link https://specs.openstack.org/openstack/docs-specs/specs/pike/os-manuals-migration.html 15:35:06 <yamahata> #link https://etherpad.openstack.org/p/doc-migration-tracking 15:35:23 <manjeets> but if I use same config locally and ping instance fip via cli it works fine 15:35:59 <mkolesni> might be a chain of events causing this 15:37:16 <yamahata> also we have intermittent failures of unit tests and functional tests. 15:37:20 <mkolesni> so you wont find it if you just try the one test 15:37:27 <yamahata> it's lower priority than tempest failure, though. 15:37:44 <mkolesni> i think functional tests of advanced services ala l2gw intermittently fails 15:37:57 <mkolesni> i rarely noticed that unit tests failed 15:38:04 <mkolesni> i.e. without a real cause 15:38:09 <yamahata> l2gw is the one I'd like to mention. 15:38:44 <yamahata> looks like we're talking patches/bugs 15:38:46 <yamahata> #topic patches/bugs 15:39:25 <yamahata> any other patches/bugs? rajivk? 15:39:42 <rajivk_> yes 15:39:59 <rajivk_> Xiao reported a few bugs 15:40:20 <rajivk_> on ocata and he wants us to merge in master well 15:40:44 <rajivk_> https://bugs.launchpad.net/networking-odl/+bug/1702417 15:41:27 <rajivk_> It was fixed after ocata release so i have raised a patch for the same 15:41:50 <yamahata> it sounds simple backport 15:41:53 <rajivk_> https://review.openstack.org/#/c/481820/ 15:42:31 <yamahata> Is it backport of two patches? 15:42:36 <yamahata> not one-by-one? 15:43:35 <rajivk_> yes, of two patches. Currently sync of resources are not supported for all the drivers, may be he has some use case of l2gw but we can plan for all resources. 15:43:37 <rajivk_> https://bugs.launchpad.net/networking-odl/+bug/1701206 15:44:20 <yamahata> If you don't have special reasons to squash patches into one, please backport patchesper upstream patch. 15:44:58 <yamahata> If you have reasons, please mention it in the commit log. 15:45:11 <manjeets> yamahata, +1 unless it is killing ci then only squash 15:45:13 <rajivk_> Ok, actually second one is to correct order of router and port. 15:45:45 <rajivk_> ok, i will raise two different patch. 15:45:48 <mkolesni> i wouldnt squash for CI sake 15:46:11 <mkolesni> it makes it harder to track in history if its squashed 15:46:40 <manjeets> mkolesni, sometimes you have to squash if two combinely go through jenkins pass 15:46:54 <rajivk_> What do you think for sync of all resources? 15:47:31 <mkolesni> oh you mean that sure but i think its rarely necessary 15:48:09 <rajivk_> i am talking about https://bugs.launchpad.net/networking-odl/+bug/1701206 15:48:43 <yamahata> rajivk_: do you mean https://review.openstack.org/#/c/478803/ ? 15:49:18 <rajivk_> yes, but it has to be raise for ocata as well 15:49:23 <rajivk_> and recovery also 15:49:36 <yamahata> After fixing it in master, we can backport patches to ocata 15:50:08 <rajivk_> okay 15:50:23 <rajivk_> https://bugs.launchpad.net/networking-odl/+bug/1701209 15:50:43 <rajivk_> I want to have your opinion on this bug. 15:51:08 <mkolesni> i dont think this is backportable since you have a db migration 15:51:32 <rajivk_> Which one, the last one? 15:51:46 <mkolesni> yes 15:52:02 <rajivk_> Do you think, it will be a good idea to fix it by other means not by db migration? 15:53:10 <mkolesni> i dunno i didnt think about it 15:53:25 <mkolesni> do you have an alternative in mind 15:53:44 <rajivk_> may be we keep a window of 5 seconds or so for retrying or not trying at exactly the same time but with time +- 5 secons 15:54:28 <rajivk_> something similar may be we can come up with better approach after discussion. 15:55:38 <mkolesni> i dont think its a huge deal 15:56:19 <mkolesni> you can set for example the timer to 301 for maintenance so only 1 in every 300 might collide :) 15:56:52 <mkolesni> but we can have some retry there with some random backoff 15:56:53 <rajivk_> hmm, may be some nearest prime number 15:57:13 <yamahata> we have 3min left 15:57:25 <yamahata> do we have any other patches/bugs? 15:57:26 <rajivk_> ok, i will put a simple patch then we can discuss over that patch. 15:57:30 <rajivk_> no 15:57:41 <mkolesni> yamahata, id like to mention https://review.openstack.org/444648 15:58:01 <yamahata> mkolesni: please go ahead 15:58:17 <mkolesni> can we get it reviewed as a stop gap? 15:58:33 <mkolesni> also then rajivk_ can continue work on thread pool 15:58:53 <yamahata> The counter patches are https://review.openstack.org/#/q/status:open+project:openstack/networking-odl+branch:master+topic:bug/1683797 15:59:00 <yamahata> Do you have any issue with them? 15:59:10 <yamahata> It makes the number of journal thread configurable. 15:59:17 <yamahata> You can set it to 1 for your case. 16:00:04 <mkolesni> i havent seen them but im not sure they collide 16:00:17 <yamahata> maybe we need to implement threadpooling in our own instead of existing threadpool. 16:00:58 <yamahata> mkolesni: okay, then please review them. 16:01:00 <mkolesni> did you see my review on https://review.openstack.org/#/c/452647/7? 16:01:10 <mkolesni> i talked with rajivk_ over it and he agrees 16:01:24 <mkolesni> so its not colliding with singleton approach 16:01:34 <mkolesni> again im not sure why the big push back against it 16:01:53 <mkolesni> we have an effective singleton of maintenance thread 16:02:06 <mkolesni> so why not have one for journal thread? 16:02:07 <yamahata> openstack is using eventlet. 16:02:24 <yamahata> at socket io, they are multiplexed. 16:02:35 <mkolesni> please see my comment on https://review.openstack.org/#/c/452647/5 regarding this 16:02:49 <mkolesni> and see what actually benefits from multithreading 16:03:35 <mkolesni> im not against multithreading we just need to do it right 16:03:41 <mkolesni> now its just a mess 16:03:52 <mkolesni> theres no thought put into it 16:03:54 <yamahata> No, we can pick journal entry by mulitple threads/processes 16:04:12 <mkolesni> how? that code is atomic across the cloud 16:04:20 <yamahata> anyway if you stick to it, you can set the number of threads to 1. 16:04:51 <mkolesni> i gtg please see my comment and think about it seriously 16:05:04 <yamahata> sure, let's continue discussion at gerrit. 16:05:10 <mkolesni> ok 16:05:16 <mkolesni> thanks guys, bye 16:05:22 <yamahata> #topic open mike 16:05:30 <yamahata> thanks every one bye 16:05:36 <manjeets> thanks 16:05:39 <yamahata> #topic cookies 16:05:44 <yamahata> #endmeeting