15:01:41 <yamahata> #startmeeting neutron_northbound 15:01:41 <odl_meetbot> Meeting started Mon Oct 2 15:01:41 2017 UTC. The chair is yamahata. Information about MeetBot at http://ci.openstack.org/meetbot.html. 15:01:41 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:01:41 <odl_meetbot> The meeting name has been set to 'neutron_northbound' 15:01:59 <yamahata> seems too slow. anyway let's start 15:02:03 <yamahata> #chair mkolesni 15:02:03 <odl_meetbot> Current chairs: mkolesni yamahata 15:02:13 <yamahata> #topic agenda bashing and roll call 15:02:17 <mkolesni> #info mkolesni 15:02:21 <yamahata> #info yamahata 15:02:22 <rajivk_> #info rajivk 15:02:35 <yamahata> #chair rajivk_ 15:02:35 <odl_meetbot> Current chairs: mkolesni rajivk_ yamahata 15:02:57 <yamahata> Do we have any topics other than usual ones? 15:03:15 <mkolesni> yes 15:03:35 <yamahata> mkolesni: what topic do you have? 15:03:50 <mkolesni> fyi we have upcoming shutdown due to holidays, i will be back on oct 15th starting from this Wed 15:04:09 <mkolesni> so 4-15 im not working 15:04:23 <yamahata> so you'll skip the meeting on Oct 9. 15:04:29 <mkolesni> yes 15:04:53 <yamahata> there is ODL DDF on Oct 9-10. so I'll also skip the meeting. 15:05:04 <yamahata> Let's cancel it next week. 15:05:08 <mkolesni> right 15:05:56 <yamahata> anything else? 15:06:01 <mkolesni> i saw today gate is broken due to zuul v3 migration 15:06:24 <mkolesni> we should focus on that before anything else 15:06:57 <yamahata> Yeah. There was an announcement on it. We need to fix it if any. 15:07:34 <mkolesni> i saw first that clones from tools/tox_install.sh should probably move toi requirements 15:07:42 <mkolesni> to* 15:07:50 <mkolesni> otherwise basic jobs fail 15:08:01 <mkolesni> also legacy jobs fail but havent looked at that yet 15:08:43 <mkolesni> so if you have time please look at it 15:08:45 <yamahata> let's discuss it after routine topics. 15:08:51 <mkolesni> if not ill try to do my best tomorrow 15:08:55 <mkolesni> ok 15:08:56 <yamahata> #topic Announcements 15:09:23 <yamahata> #info ODL DDF will take place Oct 9, 10. So the meeting next week will be skipped. 15:09:34 <yamahata> Is there any announcement? 15:10:08 <yamahata> #info openstack summit Sydney will be held November 6-8, 2017. 15:10:14 <yamahata> There will be forum sessions. 15:10:26 <yamahata> Unfortunately I won't attend it though. 15:10:34 <mkolesni> i wont be there either 15:10:46 <yamahata> any other announcement? 15:11:04 <yamahata> Oh one last annoucement 15:11:12 <yamahata> #info ODL Nitrogen has been released finally. 15:11:20 <yamahata> Now Oxygen cycle is starting. 15:11:38 <mkolesni> nice 15:11:52 <yamahata> #topic action items from last meeting 15:12:03 <yamahata> I suppose we don't have any since we didn't have meeting last week 15:12:08 <yamahata> #topic Queens/Oxygen planning 15:12:20 <yamahata> Ok, now we're discussing zuulv3 issue 15:12:30 <yamahata> #info now mkolesni is looking into it. 15:12:42 <yamahata> Also I'll also look into it. 15:13:15 <yamahata> mkolesni: any other additional info to share? 15:13:24 <mkolesni> as i said, i saw first that clones from tools/tox_install.sh should probably move to requirements 15:13:40 <mkolesni> to fix the basic jobs which are global 15:14:14 <yamahata> #link http://lists.openstack.org/pipermail/openstack-dev/2017-September/122630.html [openstack-dev] [all][infra] Zuul v3 migration update 15:14:26 <yamahata> #link http://lists.openstack.org/pipermail/openstack-dev/2017-September/122841.html [openstack-dev] [devstack] zuulv3 gate status; LIBS_FROM_GIT failures 15:15:00 <yamahata> #link https://docs.openstack.org/infra/manual/zuulv3.html Zuul v3 Migration Guide 15:15:33 <yamahata> mkolesni: thanks. 15:15:43 <mkolesni> is there a reason to clone all of them rather than have them in test-requirements for example? 15:16:17 <mkolesni> or maybe were just cloning them wrong and theres a right way to do this with zuul v3 15:17:34 <yamahata> We followed how neutron does. 15:18:01 <mkolesni> i didnt see such cloning in neutron now 15:18:04 <yamahata> Now neutron project dropped tox_install.sh. 15:18:08 <mkolesni> also their gate works 15:18:19 <mkolesni> we should probably align then 15:18:56 <yamahata> c25ece63bd8eefac4520d98ed7469ee446f4f59c change it. 15:20:36 <yamahata> mkolesni: can you take care of it? 15:20:47 <mkolesni> i will look into it tomorrow 15:20:52 <yamahata> Great. 15:21:03 <mkolesni> but as i said on wed im off to holidays 15:21:18 <mkolesni> so if i cant find it tomorrow youll have to take care of it 15:21:18 <yamahata> I see. so you have only tuesday. 15:21:22 <mkolesni> yes 15:21:27 <yamahata> Ok, got it. 15:21:29 <mkolesni> only tomorrow left 15:23:16 <yamahata> ajivk_ do you have any updates? 15:23:48 <rajivk_> yamahata, i wanted to discuss a few things 15:24:00 <yamahata> yeah, please go ahead 15:24:29 <rajivk_> During recovery, why do we need to fetch resource from db? 15:24:43 <rajivk_> if we already have a journal entry, is not it enough? 15:25:00 <mkolesni> no in recovery its failed rows 15:25:11 <mkolesni> meaning the journal entry is already out of sync 15:25:24 <mkolesni> could be even out of sync with neutron's state itself 15:25:42 <mkolesni> since failed rows dont come into account of dependency calculation 15:26:08 <rajivk_> Could you please suggest a scenario? 15:26:42 <mkolesni> yes for example delete which fails because of 404 15:26:51 <mkolesni> entry will move to failed 15:27:12 <mkolesni> sorry this actually doesnt matter 15:27:30 <mkolesni> but for example if update or create fails, then the resource was deleted from neutron 15:27:36 <mkolesni> no need to handle it anymore 15:28:00 <mkolesni> or we might need to delete it from odl if its there but not in neutron 15:28:15 <mkolesni> so its all covered by recovery 15:28:36 <rajivk_> so, update and create fails at odl side then neutron still contains the journal 15:28:43 <mkolesni> yes 15:28:50 <rajivk_> and journal contains failed entry 15:29:00 <rajivk_> so how does fetching from db is needed 15:29:10 <rajivk_> we keep on trying with the same journal entry 15:29:21 <mkolesni> but its stale 15:29:29 <rajivk_> if an update arrive at neutron side on failed resource then journal is created for the same 15:29:36 <mkolesni> the resource couldve changed 15:29:58 <rajivk_> it fails definitely due to parent resource but journal entry exist 15:30:04 <mkolesni> new entry is created but old failed entry is still there and will be picked up by the recovery 15:30:07 <rajivk_> how it could have changed 15:30:25 <mkolesni> what do you mean? 15:30:27 <rajivk_> without having journal entry for the new update? 15:30:41 <rajivk_> my questions is 15:30:42 <mkolesni> you will have it, or not, it doesnt matter 15:30:55 <rajivk_> let's take a scenario 15:31:01 <mkolesni> the fact is the failed entry is considered stale 15:31:06 <mkolesni> you cant rely on it 15:31:21 <rajivk_> How that's the point 15:31:28 <rajivk_> because we will have one more entry for update 15:31:33 <mkolesni> it doesnt matter if theres a newer entry or not 15:32:24 <mkolesni> so what about the new entry? maybe it was processed already before the recovery kicks in 15:32:25 <rajivk_> You want to say if a resource creation fails then we have stale entry at ODL side? 15:32:37 <mkolesni> no 15:32:42 <rajivk_> it is not possible because parent resource does not exist 15:32:46 <mkolesni> the journal entry, which is failed, contains stale data 15:32:48 <rajivk_> so it is bound to fail 15:33:02 <mkolesni> im not sure what youre talking about 15:33:10 <mkolesni> you have for example update network 15:33:13 <mkolesni> it fails 15:33:22 <rajivk_> okay 15:33:26 <mkolesni> then this journal entry moves to failed state 15:33:28 <mkolesni> right? 15:33:41 <rajivk_> yes 15:33:45 <mkolesni> then another update of same network, succeeds 15:34:00 <rajivk_> okay 15:34:03 <mkolesni> then this journal entry is completed, or deleted, it doesnt matter 15:34:09 <mkolesni> then recovery runs 15:34:13 <mkolesni> picks up old update 15:34:26 <mkolesni> now you want to execute it and update network to a stale state? 15:35:04 <mkolesni> you have to read the network again, you cant trust the data in the failed entry 15:35:59 <rajivk_> hmm, got it. 15:36:33 <rajivk_> because we don't have good way to find among two entry, which to apply first and then second 15:36:51 <mkolesni> yes 15:37:33 <rajivk_> Got it. i am getting a lot of ifs now in the recovery and full sync 15:38:04 <mkolesni> a lot of ifs? 15:38:55 <rajivk_> Different drivers of ODL are implemented with some difference then the way neutron understand. 15:39:08 <mkolesni> ok 15:39:37 <rajivk_> for example, l2gateway-connection is used and neutron uses l2-gateway-connection 15:39:56 <rajivk_> neutron uses different attribute to store gatway id and odl understands different 15:40:13 <rajivk_> bgpvpn driver we don't have network assoc and router assoc in odl 15:40:18 <mkolesni> ok 15:40:19 <rajivk_> but they are update operations on odl 15:40:20 <yamahata> Do you mean inconsistency of naming between odl and neutron? 15:40:26 <rajivk_> yes 15:40:36 <rajivk_> and attributes name as well 15:40:49 <rajivk_> and the way neutron and odl thinks about the resources 15:41:09 <rajivk_> driver handle it, we should not handle them in recovery and full sync again 15:41:54 <rajivk_> what if for create during recovery, i use precommit of driver, it does all manipulation, i am thinking about it and complexity 15:42:03 <rajivk_> just wanted to have your opinion 15:43:09 <mkolesni> you have a poc for this? 15:43:27 <mkolesni> if its simpler go ahead 15:43:35 <mkolesni> i suspect it wont be but maybe it will 15:43:49 <rajivk_> not yet. 15:44:10 <mkolesni> ok 15:44:24 <rajivk_> yamahata, can you think of complexity or the problem we can encounter? 15:44:39 <yamahata> right now I don't see big one. 15:44:55 <yamahata> If you have poc patch, we can discuss it based on poc. 15:45:38 <yamahata> In ideal world, there shouldn't be naming inconsistency. In reality we can't avoid it given the current situation. 15:45:42 <rajivk_> okay, i will try a PoC patch for it. 15:46:09 <yamahata> some adaptation layer would be inevitable probably. 15:46:25 <rajivk_> between which two layers 15:47:52 <rajivk_> regarding this https://review.openstack.org/#/c/504297/ 15:48:14 <yamahata> new lib/neutron patch? 15:48:26 <rajivk_> yamahata, yes, do we want to keep neutron legacy and neutron both together? 15:48:44 <yamahata> Eventually neutron legacy will be removed. 15:48:53 <yamahata> During transition we would need to keep both. 15:49:20 <rajivk_> okay 15:49:39 <rajivk_> now, their are many workaround need to cleanup it, once gate job are working 15:49:40 <yamahata> Once we're confident with new lib/neutron, we can remove legacy one. 15:50:39 <rajivk_> okay, that's all, i wanted to discuss 15:50:46 <yamahata> Great. 15:50:59 <yamahata> From me, my concern is ODL 404 bug. 15:51:08 <yamahata> I'm not sure anyone is actively working on it or not. 15:51:16 <mkolesni> not that i know 15:51:53 <yamahata> Okay. that's unfortunate. 15:52:03 <yamahata> and we have new 5 bug report since PTG. 15:52:09 <yamahata> 2 are already in progress. 15:52:37 <yamahata> that's all I wanted to share. 15:52:46 <yamahata> #topic open mike 15:52:48 <mkolesni> ok 15:52:51 <yamahata> anything else? 15:52:56 <yamahata> otherwise I'll give you back 8mins 15:53:13 <mkolesni> btw next week mpeterson should be working Sun-Tue so you guys can ping him if you need something 15:53:30 <yamahata> okay. 15:53:40 <yamahata> Thank you everyone. 15:54:00 <rajivk_> thanks Mike, Yamahata 15:54:02 <yamahata> #topic cookies 15:54:07 <yamahata> #endmeeting