15:11:49 <yamahata> #startmeeting neutron_northbound 15:11:49 <odl_meetbot> Meeting started Mon Jun 12 15:11:49 2017 UTC. The chair is yamahata. Information about MeetBot at http://ci.openstack.org/meetbot.html. 15:11:49 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:11:49 <odl_meetbot> The meeting name has been set to 'neutron_northbound' 15:11:51 <yamahata> hmm, bot doesn't seem to be working. 15:11:53 <yamahata> #topic agenda bashing and roll call 15:11:55 <yamahata> #chair mkolesni 15:11:55 <odl_meetbot> Current chairs: mkolesni yamahata 15:11:55 <jhershbe> yamahata, we were all here 15:11:57 <yamahata> anyone else? 15:11:59 <yamahata> mkolesni: yes. 15:12:01 <yamahata> #chair jhershbe 15:12:01 <odl_meetbot> Current chairs: jhershbe mkolesni yamahata 15:12:07 <jhershbe> me too 15:12:09 <mkolesni> #info mkolesni 15:12:11 <yamahata> #info yamahata 15:12:18 <jhershbe> #info jhershbe 15:12:27 <yamahata> any other topic in addition to usual topics? 15:12:46 <jhershbe> odl_features and port status patches 15:13:03 <yamahata> it can be discuassed at usual patches/bugs 15:13:13 <yamahata> #topic Announcements 15:13:27 <mkolesni> i wanted to discuss a patch that was merged 15:13:31 <yamahata> pike-2 was reached and it was released. 15:13:39 <yamahata> mkolesni: sure. 15:13:41 <mkolesni> https://review.openstack.org/#/c/459970/15 15:14:05 <mkolesni> pike-3 is feature freeze right? 15:14:10 <mkolesni> 696187 15:14:40 <yamahata> #link https://releases.openstack.org/pike/schedule.html release schedule 15:14:54 <mkolesni> thats what i understand from the schedule 15:15:10 <yamahata> Yes, pike-3 is feature freeze on Jul 24 - Jul 28 15:15:33 <yamahata> Now CI isn't in good shape, so it needs to be fixed. 15:15:37 <mkolesni> ok let me know when we can discuss that patch 15:15:50 <mkolesni> whats the problem with ci? 15:15:57 <yamahata> mostly floatingip, grenade, fullstack, functional aren't working well. 15:16:01 <yamahata> They are mostly failing. 15:16:26 <mkolesni> its all with new netvirt? 15:16:31 <yamahata> Right. 15:16:39 <yamahata> and v2driver failing rate is wrose than v1driver. 15:16:45 <yamahata> #link http://grafana.openstack.org/dashboard/db/networking-odl-failure-rate 15:17:27 <yamahata> on ODL side, karaf4 migration is planned on Nitrogen cycle. 15:18:00 <yamahata> right now ODL NN build is broken. it also need fix. 15:18:09 <yamahata> any other announcement? 15:19:22 <mkolesni> move on? 15:19:28 <yamahata> seems nothing, let's move on 15:19:33 <yamahata> #topic action items from last meeting 15:19:43 <yamahata> so far except patch review etc, there isn't specific one. 15:19:58 <yamahata> #topic Pike/Nitrogen planning 15:20:17 <yamahata> Now Pike cycle is begining. and ODL NN is applying for ODL simultenious release. 15:20:21 <yamahata> It's M-0 15:20:32 <yamahata> #topic patches/bugs 15:20:38 <mkolesni> can we talk about the aforementioned patch? 15:20:52 <yamahata> sure, mkolesni or jhershbe which is first? 15:21:01 <mkolesni> mine is shorter i guess 15:21:10 <mkolesni> jhershbe, you mind? 15:21:16 <jhershbe> mkolesni, not at all 15:21:20 <yamahata> mkolesni: okay, https://review.openstack.org/#/c/459970/ 15:21:24 <mkolesni> #link https://review.openstack.org/#/c/459970/15 15:21:39 <mkolesni> ok it was merged about 10 days ago i noticed it last week 15:21:44 <mkolesni> i put some comments there 15:21:46 <yamahata> I'm seeing your comment. 15:21:54 <mkolesni> basically its quite problematic 15:22:15 <mkolesni> file was renamed but not using git mv so we lose ability to track it 15:22:44 <mkolesni> also the db directory should be pike not ocata as ocata was released a while back 15:23:09 <mkolesni> also i think just brute force trying to create the row lock is bad design 15:24:07 <mkolesni> we inserted the row permanently on db table creation to avoid races and also this "lazy" code which is quite unnecessary since this row's existence is constant 15:24:19 <mkolesni> also theres a bug there with row unlocking 15:24:33 <mkolesni> so generally many problems that were introduced with that patch 15:24:54 <mkolesni> now we have 2 options, either fix it with more patches or revert that one and re introduce it properly 15:25:18 <yamahata> either way is okay with me. 15:25:25 <yamahata> mkolesni: can you take care of it? 15:25:28 <jhershbe> I vote for revert and re-introduce so the history of hte file is preserved 15:25:37 <mkolesni> ok sure ill try 15:25:59 <mkolesni> im not sure if the history can be preserved but ill try a method from the internets 15:26:01 <yamahata> onething, git mv is only convenient command, git remove/add. so it doesn't matter. 15:26:21 <mkolesni> not quite it also knows to track the history 15:26:31 <yamahata> please see git log --find-renames options. 15:26:42 <yamahata> and unfortunately gerrit isn't very good at handling rename. 15:27:33 <yamahata> git doesn't track rename directly. It guesses rename by contents similarity. 15:27:51 <mkolesni> ok nvm i see git does eat it well so maybe its just the way gerrit displays it, but the other points are valid 15:27:58 <mkolesni> so ill take care of it tomorrow then 15:28:14 <mkolesni> jhershbe, the floor is yours 15:28:27 <jhershbe> yamahata, so let's start with odl_features... 15:28:49 <yamahata> jhershbe: sure. 15:29:16 <jhershbe> yamahata, as I stated, I'm not happy with the "abort if odl not available" option. I put some responses in gerrit to your last comment. Probably best if we all start from that 15:29:16 <yamahata> #link https://review.openstack.org/#/c/465462/ odl_features patch 15:29:56 * yamahata is reading the comment 15:30:14 * jhershbe is waiting patiently ;-) 15:33:03 <mkolesni> the comment is quite lengthy :) 15:33:04 <yamahata> okay, basically you'd like to stick to the logic of patchset 29-? 15:33:17 <jhershbe> yamahata, preferably. 15:34:56 <yamahata> So why you don't like boot up abort or forver loop? 15:35:04 <mkolesni> i have to agree with jhershbe 15:35:21 <mkolesni> its a production grade piece of software 15:35:24 <jhershbe> abort is the worst option, IMHO. What do you gain by not looping forever and waiting for ODL? 15:35:24 <yamahata> You said there is no advantage. or disadvantage? 15:35:32 <mkolesni> as such it has to be as most resilient as we can make it 15:36:17 <jhershbe> Yes. Resilience. 15:36:33 <jhershbe> PS29 gives us max resilience now 15:36:36 <yamahata> with forever loop, the logic will be simple and we don't have to worry about the use of has_feature. 15:36:47 <jhershbe> mkolesni, ps29 is the async model 15:36:50 <mkolesni> we cant shutdown neutron server or even the driver just because ODL is momentarily dead 15:37:01 <mkolesni> jhershbe, yes i understand from the context 15:37:53 <jhershbe> yamahata, I think looping forever is _acceptable_ 15:37:55 <mkolesni> jhershbe, though IMO the init() itself should be done in a thread 15:38:27 <jhershbe> mkolesni, I also think so but yamahata is worried about features changing at runtime which can happen in that case 15:38:42 <mkolesni> then lets tackle that problem when it happens 15:38:58 <jhershbe> mkolesni, couldn't agree more. yamahata ? 15:39:01 <yamahata> another option would be, we can write a small script to setup features_set by retrieaving running ODL. 15:39:15 <yamahata> then put it into neutron configuration. 15:39:30 <yamahata> so that networking-odl can boot without ODL booting. 15:39:47 <jhershbe> not understanding 15:39:47 <yamahata> It requires cloud admin run ODL once and configure neutron, though. 15:40:58 <yamahata> mkolesni: why would you like to run thread for init() 15:41:22 <jhershbe> yamahata, I have a proposal... 15:41:58 <yamahata> jhershbe: yes? what's your proposal? 15:42:06 <jhershbe> we do loop forever *unless* there is local config for the features, in which case we use those. AND we provide the script 15:42:32 <jhershbe> yamahata, that way people will not need to know about the script 15:43:00 <jhershbe> like people who are just running devstack or something 15:43:20 <jhershbe> I don't like the idea of adding a setup step that is required for all cases. 15:43:35 <jhershbe> In other words... 15:44:13 <jhershbe> For convenience, for most cases, especially development, the fetch on init (and loop forever) will work just fine... 15:44:31 <jhershbe> if someone wants enterprise grade, they run the script or populate the features in the config themselves. 15:44:47 <jhershbe> Best, and simplest of all worlds 15:45:09 <mkolesni> im actually not against that option as an optimization to this approach 15:45:25 <mkolesni> sounds good generally 15:45:30 <yamahata> that sounds great. 15:45:41 <mkolesni> so we have the config and if its not set then we poll odl endlessly, right? 15:45:58 <jhershbe> ok 15:46:06 <mkolesni> sounds good to me 15:46:23 <jhershbe> So, how does this sound... 15:46:47 <jhershbe> yamahata, I will update the patch with that code. 15:47:04 <yamahata> jhershbe: thank you. glad to reach consensus. 15:47:15 <jhershbe> yamahata, but I do want to talk to my boss first but I assume we can agree on it 15:47:39 <jhershbe> yamahata, I'd like to discuss your final comment on https://review.openstack.org/#/c/465463/ as well 15:48:10 <jhershbe> I'd like to postpone the lock till after these are merged as a separate patch 15:48:10 <yamahata> sure. 15:48:16 <jhershbe> sound good? 15:48:42 <yamahata> Yes. as a separate patch. 15:49:09 <jhershbe> ok, cool. So I will push a new PS for the odl_features tomorrow or the day after 15:49:27 <mkolesni> yamahata, can you please review https://review.openstack.org/#/c/444648/12 ? 15:49:38 <jhershbe> guys I'm outa here 15:49:50 <mkolesni> yamahata, we discussed last week but didnt see a review from you 15:49:53 <yamahata> sure. 15:49:56 <mkolesni> ok thanks 15:50:02 <mkolesni> so i have nothing further 15:50:07 <mkolesni> i think rajivk had something 15:50:14 <yamahata> rajivk: are you there? 15:50:21 <rajivk> yes 15:50:35 <rajivk> mkolesni already discussed about it 15:50:37 <yamahata> do you have anything to discuss? any patches/reviews? 15:50:57 <yamahata> okay. 15:50:57 <rajivk> no 15:51:08 <yamahata> #topic open mike 15:51:12 <yamahata> anything else to discuss? 15:51:16 <mkolesni> :) 15:51:22 <mkolesni> nothing from me 15:51:49 <yamahata> we're done. 15:51:53 <yamahata> thank you everyone. 15:52:03 <yamahata> #topic cookies 15:52:08 <yamahata> #endmeeting