15:01:20 <yamahata> #startmeeting neutron_northbound
15:01:20 <odl_meetbot> Meeting started Mon Jan  8 15:01:20 2018 UTC.  The chair is yamahata. Information about MeetBot at http://ci.openstack.org/meetbot.html.
15:01:20 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:01:20 <odl_meetbot> The meeting name has been set to 'neutron_northbound'
15:01:35 <yamahata> #chair mkolesni mpeterson
15:01:35 <odl_meetbot> Current chairs: mkolesni mpeterson yamahata
15:01:42 <yamahata> #topic agenda bashing and roll call
15:01:47 <yamahata> #info yamahata
15:01:59 <yamahata> any additional topics?
15:02:01 <manjeet> #info manjeets
15:02:07 <yamahata> #chair manjeet
15:02:07 <odl_meetbot> Current chairs: manjeet mkolesni mpeterson yamahata
15:02:07 <mkolesni> #info mkolesni
15:02:10 <mpeterson> #info mpeterson
15:02:25 <yamahata> I know I'm behind patch review and catching them up.
15:03:02 <mkolesni> i wanted to discuss l3 flavor
15:03:39 <mkolesni> #info discuss l3 flavor
15:03:44 <yamahata> okay.
15:03:48 <yamahata> any other topics?
15:04:14 <mpeterson> not here
15:05:57 <yamahata> okay.
15:06:02 <yamahata> move on
15:06:07 <yamahata> #topic Announcements
15:06:40 <yamahata> Jan 22-26 is queens Milestone-3. feature freeze
15:07:10 <yamahata> Jan 7 is ODL M-4
15:07:29 <yamahata> oops Jan 14 is ODL M-4 for offset -1
15:07:38 <mpeterson> #info Jan 22-26 is queens Milestone-3. feature freeze
15:07:45 <yamahata> it's code freeze. Code Freeze
15:07:50 <yamahata> any other announcement?
15:08:08 <mpeterson> #info Jan 14 is ODL M-4 for offset -1 code freeze
15:09:16 <mkolesni> do you know if you'll be going to ptg yet?
15:09:33 <yamahata> Unfortunately not yet.
15:09:44 <mkolesni> that is unfortunate
15:10:43 <yamahata> any way this week or next week I need to get outcome for travel arrangement.
15:10:54 <mkolesni> ok cool
15:11:17 <yamahata> #topic action items from last meeting
15:11:28 <yamahata> there was no action items.
15:11:36 <yamahata> #topic patches/bugs
15:11:45 <yamahata> okay, any patches/bugs to discuss?
15:11:50 <yamahata> l3 flavor first?
15:12:13 <yamahata> mkolesni: you're on stage.
15:12:14 <mkolesni> sure should be quick
15:12:17 <mkolesni> ok
15:12:25 <mkolesni> so i was looking at the l3 flavor patch
15:12:40 <mkolesni> and then i was looking at l3_db.py in neutron
15:13:03 <mkolesni> then i came to a realization, that we now have precommit registry calls for routers and floating ips
15:13:09 <mkolesni> right?
15:13:28 <yamahata> the patch for neutron l3 isn't merged yet.
15:13:39 <yamahata> but, we will have them.
15:13:41 <mkolesni> no im looking at neutron master
15:14:24 <mkolesni> look for example at https://github.com/openstack/neutron/blob/master/neutron/db/l3_db.py#L228
15:14:34 <yamahata> I mean https://review.openstack.org/#/c/523257/ . are you talking about different patch?
15:14:41 <yamahata> oh, let me check it.
15:15:12 <mkolesni> in the neutron code base, there's already calls to notify on precommit or router/fip CUD operations
15:15:42 <yamahata> It's already used by driver_controller.py. So yet another callbacks are needed.
15:16:03 <mkolesni> so given this, do we even care about l3 flavor? why not have odl-l3 "service plugin" just register to these calls?
15:16:24 <mkolesni> not sure what do you mean here, whats missing?
15:16:29 <yamahata> we care l3 flavor.
15:17:14 <yamahata> other l3 can be active in addition to odl.
15:17:20 <yamahata> So l3 flavor is needed.
15:17:27 <mkolesni> the code in neutron is already calling these notify anyway, so anything in the registry will get those and can act upon it right?
15:17:44 <mkolesni> well right now the situation is monolithic plugin
15:18:00 <mkolesni> im talking about the fact that current l3 is quite broken
15:18:16 <yamahata> Do you mean l3 = odl l3?
15:18:23 <mkolesni> yes
15:18:39 <mkolesni> we could convert to use the registry calls already now in the monolithic plugin
15:18:47 <manjeet> I guess that could be possible if we can register odl l3 as a flavor driver
15:18:58 <manjeet> and use the registry calls inside it
15:19:07 <mkolesni> later on when l3 flavors are introduced this can be adapted or rewritten to use that
15:19:38 <mkolesni> ok a clarification then
15:19:40 <yamahata> Now we have already l3 flarvor patch for neutron and odl l3 flavor patch under review.
15:19:51 <mkolesni> im talking about this code currently - https://github.com/openstack/networking-odl/blob/master/networking_odl/l3/l3_odl_v2.py
15:19:59 <yamahata> that is the plan for queens.
15:20:05 <mkolesni> this implements sort of an l3 plugin right?
15:20:25 <yamahata> so we can replace odl_l3_v2 by odl l3 flavor driver.
15:20:40 <mkolesni> so instead or overwriting the methods and inheriting l3 plugin, why not change it to receive the registry calls?
15:21:05 <mkolesni> yes i understand but FF is 2 weeks away and i don't think flavors will work by then
15:21:15 <mkolesni> we need a working solution that's better than what we have
15:21:42 <mkolesni> im not saying dump l3 flavors, im saying we should fix the current code to work better
15:21:50 <mkolesni> to make in time for ff
15:22:23 <yamahata> we can apply FF exception.
15:22:37 <yamahata> we can talk with yamamoto.
15:23:04 <mkolesni> ok but its quite risky dont you want at least the monolithic plugin we have to work properly?
15:23:19 <yamahata> mkolesni: btw why are you suddenly concorned about l3?
15:23:41 <yamahata> so far you're very ignorant about broken odl l3 driver and had blocked efforts to fix it.
15:23:59 <mkolesni> it seems to me l3 flavors wont be ready (i.e, production grade) in time for ff
15:24:16 <mkolesni> we faced some bugs with l3 on our scale testing, so im taking an interest in it
15:25:13 <yamahata> gotcha.
15:25:26 <manjeet> mkolesni, are those bugs reported to launchpad ?
15:25:37 <manjeet> or not yet ?
15:25:39 <mkolesni> manjeet, yes i asked them to report the bugs
15:25:47 <yamahata> I don't see any point to fix odl l3, l3 flavor should be focused.
15:26:16 <yamahata> neutron part is mostly ready as it got 1 +2.
15:26:16 <manjeet> ohk mkolesni may be you can notify once those are reported
15:26:25 <yamahata> remaining part is update.
15:26:26 <mkolesni> ok so if i fix the current l3 you won't approve it?
15:26:50 <yamahata> I don't see any point for odl l3 plugin. we have had ODL l3 plugin broken for long time.
15:27:32 <yamahata> you've been against fixing odl l3 plugin for long time.
15:27:36 <mkolesni> you can't remove it right now since flavors is something different
15:27:53 <mkolesni> i havent been against fixing it, i disagreed with the proposed solution
15:28:13 <mkolesni> if what im talking about now was proposed i'd be happy with it
15:28:26 <mkolesni> but the solution that was in place was a workaround not a fix
15:29:42 <yamahata> Right. l3 flavor is the solution.
15:30:04 <yamahata> (router, precommit) callbacks is already used by driver controller.
15:30:22 <mkolesni> im not sure what you mean by that
15:31:03 <poothia_> #info poothia
15:31:15 <mkolesni> registry.notify calls all the registered callbacks
15:31:22 <mkolesni> so if we register for it we'll get this also
15:31:38 <yamahata> it looks like this topic will take time.
15:31:39 <mkolesni> same like we do for sec groups
15:31:56 <yamahata> any other patches/bugs to discuss? and then we can back to l3 flavor discussion.
15:33:16 <yamahata> If no, we can just continue l3 flavor discussion.
15:33:55 <yamahata> seems no other topic. let's continue l3 flavor discussion.
15:34:33 <mkolesni> so if i fix l3_odl_v2 to work with the registry callbacks, you're saying you'll -2 it?
15:35:04 <yamahata> Maybe.
15:35:17 <mkolesni> what do you mean?
15:35:22 <yamahata> because l3 flavor is solution, we should focus on l3 flavor.
15:35:33 <mkolesni> but this isnt related
15:35:52 <mkolesni> lets consider the possibility that l3 flavor wont be stable in queens
15:36:01 <mkolesni> why not have something stable we can use?
15:36:33 <yamahata> In that case, the stabilization will be in Rocky cycle.
15:37:00 <mpeterson> just to fill my ignorance, are l3 flavors stable from a neutron perspective? are they a workable solution?
15:37:19 <mkolesni> fine but why not ship a stable l3 solution in queens?
15:37:39 <yamahata> we can try to stabilize l3 flavor in queens.
15:37:57 <mkolesni> we cant count on this since its not stable in neutron yet
15:38:06 <mkolesni> and stabilizing in neutron is much harder
15:39:07 <mkolesni> look if you're going to -2 it just say it now so we dont waste time trying to fix it
15:39:16 <yamahata> then we should unit to stabilize neutron part because it's hard.
15:39:54 <mkolesni> l3 flavors was supposed to be usable in Pike but it slipped that, right?
15:40:16 <mkolesni> so you can't guarantee it will be usable in Queens
15:40:28 <mkolesni> or even in Rocky for that matter
15:40:32 <yamahata> you can also help to make it in queens.
15:40:55 <mkolesni> so why not fix the current l3 we have to work properly?
15:41:01 <mkolesni> why are you so against it?
15:41:16 <yamahata> l3 flavor will replace odl l3 plugin and l3 flavor is the solutions.
15:41:22 <yamahata> Let's make it in queens.
15:41:40 <mpeterson> yamahata, mkolesni: could you please answer my question please?
15:42:48 <yamahata> In neutron part, it's been workable for ovs case. for ODL, one more patch is needed. https://review.openstack.org/#/c/523257/
15:43:26 <yamahata> l3 flavor framework has been tested for more than one cycles.
15:44:09 <mpeterson> yamahata: thanks for replying
15:44:11 <mkolesni> just like we still have several port binding controllers, we still need the monolithic l3 plugin in case l3 flavors dont work well
15:44:31 <mkolesni> im just proposing to fix this monolithic plugin to work more properly
15:44:38 <mkolesni> as a contingency plan
15:45:08 <yamahata> so far we have l3 plugin broken for long time, I don't think it's needed.
15:45:12 <yamahata> just we can focus l3 flavor.
15:45:16 <mkolesni> and obviously l3 flavors was tested for several cycles with only one actual flavor, so is it really well tested?
15:45:25 <mkolesni> i dont think so
15:45:58 <yamahata> maybe or maybe not. we can focus on stabilizing l3 flavor.
15:46:09 <yamahata> neutron or networking-odl.
15:46:53 <mpeterson> yamahata: can we agree to allow a solution to the monolithic plugin to be considered as well as it became relevant for us now?
15:47:21 <mkolesni> look we (red hat) need a working solution for queens, i suggest we fix the monolithic plugin and in the meanwhile work can also continue on l3 flavors
15:47:30 <yamahata> so far we have focused on l3 flavor. I don't like distraction.
15:47:52 <mpeterson> yamahata: right, but how do  we address our companies needs if you won't accept a backup plan?
15:48:08 <mpeterson> yamahata: it's not like the work on the monolithic plugin goes against the flavor
15:48:23 <mpeterson> yamahata: they are two different solutions that address the same issue
15:48:41 <yamahata> Hmm, have you reviewed any patches related to l3 flavor?
15:49:11 <yamahata> Given that history, I really don't want distraction.
15:49:18 <mkolesni> yes i looked at it and it seems like quite a complicated mechanism right now
15:49:22 <mpeterson> yamahata: I have not. What would that change?
15:51:38 <yamahata> I'm very concerned that you fix l3 plugin and then go for other place and l3 flavor will be left untouched.
15:52:35 <mpeterson> yamahata: if that's where the overarching project is going then it will lead us to l3 flavors in the end
15:52:59 <yamahata> what's the deadline for our decision of l3 flavor?
15:53:17 <mkolesni> im waiting for the code to land in neutron since it will change the n-odl implementation
15:53:32 <mkolesni> it needs to land before feature freeze right?
15:53:37 <yamahata> Q-3? Jan 22-26?
15:54:14 <mkolesni> yes thats what feature freeze means afaik
15:54:19 <yamahata> yeah, after that we can apply FF exception.
15:54:45 <yamahata> FFE would be risky bet.
15:55:37 <mkolesni> they might not even grant it
15:56:03 <mkolesni> were not going to ignore l3 flavors but we need something working right now
15:57:24 <mkolesni> problem is that whats there right now is quite buggy
15:58:01 <yamahata> that's not surprize. the bug has been known for long time.
15:58:23 <yamahata> okay, anyway we need to come up with conclusion.
15:58:44 <yamahata> does next week decision work for you?
15:59:12 <mkolesni> if you're not going to -2 it id like to start work right away so that we have time to test it
15:59:33 <mkolesni> the fix to existing code
16:00:04 <yamahata> my personal bet is to focus on l3 flavor, though.
16:00:09 <yamahata> anyway we'll see outcome.
16:00:19 <yamahata> anything else to discuss?
16:00:26 <yamahata> we're runinning out of time.
16:00:43 <mpeterson> yamahata: I don't understand why are you refusing this so harshly
16:01:08 <yamahata> I'm concerned about igrance of l3 flavor.
16:01:11 <mpeterson> yamahata: mkolesni is asking for a backup plan that addresses the needs our clients have
16:01:20 <yamahata> it's distract.
16:01:32 <mpeterson> yamahata: l3 is not ready yet
16:01:44 <mpeterson> yamahata: and we don't have a workable ETA of when it will be ready
16:01:54 <mkolesni> if l3 flavor would be ready i'd already review it
16:02:05 <mpeterson> yamahata: and there is a need that can be addressed in the short term with this backup plan
16:02:20 <mkolesni> problem is it isn't and doesn't look like it's going to be in time for FF
16:02:24 <mkolesni> thats my concern
16:02:26 <yamahata> mkolesni: the mind set is wrong. we should make l3 flavor ready.
16:02:32 <yamahata> not wait for its readiness.
16:03:07 <mpeterson> yamahata: I understand your concerns, but we are a community that should work together, not against each other.
16:03:16 <mkolesni> afaik enough people including you are working on it, i dont see how it will make it more ready
16:03:37 <mkolesni> anyway i gtg
16:03:45 <yamahata> okay. thank you every one.
16:03:47 <mkolesni> ill keep the discussion on the email
16:03:53 <mpeterson> yamahata: by taking this decision your a knowingly hurting the concern mkolesni is presenting, which in turn hurts the company concerns
16:05:59 <yamahata> #topic cookies
16:06:04 <yamahata> #endmeeting