15:00:42 <mpeterson> #startmeeting weekly meeting 15:00:42 <odl_meetbot> Meeting started Mon Apr 16 15:00:42 2018 UTC. The chair is mpeterson. Information about MeetBot at http://ci.openstack.org/meetbot.html. 15:00:42 <odl_meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:42 <odl_meetbot> The meeting name has been set to 'weekly_meeting' 15:00:53 <mpeterson> #topic Agenda 15:00:55 <mpeterson> Any topics to discuss? Topics I have are: action items from last meeting, announcements, patches 15:00:56 <mkolesni> is it possible to move this 1 hour earlier? 15:01:03 <poothia__> #info poothia 15:01:23 <mpeterson> poothia__: there is no need to issue an #info, the meetbot knows yu are here if you talk 15:01:36 <poothia__> ok... :) 15:01:47 <manjeets> mkolesni, I have no issues with moving it an hour early 15:02:03 <mkolesni> yamahata, ? 15:02:12 <mpeterson> poothia__: you too are ok with the meeting moving 1 hour earlier? 15:02:13 <rajivk> mkolesni, may be we can move it by 30 minutes. Is that ok? 15:02:38 <mpeterson> rajivk: it's actually a bit late already here... 1 hour is difficult for you? 15:02:44 <mkolesni> rajivk, 1 hour is better, if we cant then it can stay as is i guess 15:03:16 <rajivk> mkolesni, mpeterson, i might miss sometimes if we make it one hours earlier 15:03:37 <mpeterson> mmm then I think it's better to leave it as is, what do you think mkolesni ? 15:03:55 <mkolesni> ok but i might miss it from time to time 15:04:17 <rajivk> mkolesni, i can give you 15 more minutes. 15:04:24 <mpeterson> oh... then it's a complicated situation 15:04:36 <mkolesni> no lets leave it like this then 15:04:54 <mkolesni> well see how this goes 15:04:58 <mpeterson> okey, any topics other than the ones I said? 15:05:03 <mpeterson> so we can move forward 15:05:25 <manjeets> I guess since mkolesni is present today we can go over bug list later 15:05:37 <mkolesni> ok 15:05:43 <mpeterson> okey, I'll add it and we'll see how the time goes 15:06:41 <mpeterson> #info added topics: v3 driver, bug triage 15:06:48 <mpeterson> ##topic Action items from last meeting 15:06:49 <mpeterson> #topic Action items from last meeting 15:07:06 <mpeterson> okey, so we have some outstanding actions 15:07:20 <mpeterson> #action mkolesni, yamahata, manjeets to review Ml2 for new full sync 15:07:35 <mpeterson> #action manjeets to review https://review.openstack.org/#/c/529537/ 15:07:53 <mpeterson> please try to do those for next time, so we can move forward 15:08:06 <mkolesni> this is a short week here btw, only 1 work day left 15:08:20 <mpeterson> mkolesni: true that, perhaps the rest can do more 15:08:32 <manjeets> mpeterson, I'll out of office from tomorrow for a conference, but will try for sure to get those done 15:08:43 <mpeterson> manjeets: thanks 15:08:46 <mpeterson> #topic Announcements 15:09:15 <mpeterson> anyone has an announcement? nothing on my part and the PTL is not here :( 15:10:17 <mpeterson> I'm going to assume that's a no 15:10:17 <manjeets> I may need reviews on l3 flavors driver, but waiting on neutron-lib release to unbreak patch in neutron which is needed for n-odl 15:10:35 <manjeets> but code reviews can still help 15:10:44 <mpeterson> #info no announcements 15:10:53 <mpeterson> #topic Patches 15:11:16 <mpeterson> okey, so what's the review topic/patch that you want us to review? 15:11:22 <mpeterson> manjeets: ^^ 15:11:37 <manjeets> https://review.openstack.org/#/c/504182/ 15:11:44 <manjeets> https://review.openstack.org/#/c/544116/2 15:11:50 <manjeets> functional tests are also coming 15:12:35 <mpeterson> #action all to review L3 ( https://review.openstack.org/#/c/504182/ and https://review.openstack.org/#/c/544116/2 ) 15:12:47 <mpeterson> manjeets: I don't think I'll be able to review it before the next meeting but I'll do my best 15:12:58 <manjeets> mpeterson, np thanks ! 15:13:26 <mpeterson> on my part, I'd need people to start reviewing the implementation of the enginefacade in ODL 15:13:48 <mpeterson> #action all to review bp/enginefacade-switch ( https://review.openstack.org/#/q/status:open+project:openstack/networking-odl+branch:master+topic:bp/enginefacade-switch ) 15:14:19 <mpeterson> anything else on patches? 15:15:20 <mpeterson> moving to the next topic in 20 seconds 15:15:32 <rajivk> yes 15:15:48 <rajivk> it would be great if your guys could review https://review.openstack.org/#/c/558554/ 15:16:11 <rajivk> and of course full sync and recovery one https://review.openstack.org/#/c/527391/ 15:16:41 <mpeterson> rajivk: yes, they are in the todo list already, on actions from last meeting 15:17:10 <mpeterson> oh, the first one isn't 15:17:37 <mpeterson> #action all to review the feature negotiation patch https://review.openstack.org/#/c/558554/ 15:18:05 <mpeterson> a lot of work to do it seems 15:18:16 <mpeterson> okey, moving on 15:18:18 <mpeterson> #topic v3 driver 15:18:19 <mkolesni> its a good thing :) 15:18:27 <mpeterson> mkolesni: indeed 15:19:07 <manjeets> are you guys coming with new v3 drivers ? 15:19:16 <mpeterson> as it was brought mkolesni during the PTG, networking-ovn has a better solution to what our journal triest ot solve 15:19:16 <manjeets> comin up** 15:19:39 <mpeterson> this solution is based on the revision number that's present in neutron's standard attributes 15:19:42 <mkolesni> manjeets, didnt isaku update about that? 15:19:52 <manjeets> oh yes there's note on etherpad too 15:20:00 <mkolesni> ok 15:20:04 <mpeterson> and it's quite simple and nice 15:20:08 <manjeets> he did update me but it was out of my mind atm 15:20:14 <mkolesni> :) 15:20:29 <manjeets> ok, I actually need to look at how n-ovn does it to get more details 15:20:58 <mpeterson> I'm working on a spec for the integration in neutron 15:21:12 <mpeterson> so several drivers can leverage from one implementation 15:21:12 <mkolesni> mpeterson, can u please post the link to ovn spec? 15:21:23 <mpeterson> but for a rough idea you can read the ovn spec https://github.com/openstack/networking-ovn/blob/master/doc/source/contributor/design/database_consistency.rst 15:21:28 <mkolesni> so that everyone can later read to get the context 15:21:35 <mkolesni> mpeterson, 10x 15:21:56 <mpeterson> #info we are looking to implement OVN's spec in neutron in a more generic way, mpeterson is working on it (ovn implementation: https://github.com/openstack/networking-ovn/blob/master/doc/source/contributor/design/database_consistency.rst ) 15:23:29 <mpeterson> by implementing this idea also we get to do full_sync in a much easier way, and also recovery might not be needed at all 15:24:16 <mpeterson> #action all make yourselves familiar with the design and if you have any ideas/questions/contributions, you can contact me so I can included/answer them 15:24:22 <mkolesni> there are many benefits 15:24:35 <mkolesni> do you want me to list the ones i have in mind? 15:24:44 <mpeterson> mkolesni: yes, that would be nice 15:24:51 <mkolesni> ok 15:25:20 <mkolesni> first, obviously, shared code increases maintainability since we'll have more eyes on it to fix any bugs or add new features 15:25:26 <mpeterson> #info mkolesni will list many of the benefits, you can read the irclogs of the meeting to have an idea 15:25:48 <mkolesni> second, the solution is quite nice and elegant so it is much more maintainable by itself 15:26:20 <manjeets> mkolesni, sharing code with networking-ovn right (is neutron-lib gonna be home for all the common code ?) 15:26:23 <mkolesni> also since it is shared and separated from our core driver code, we can focus on the core aspects of integration with odl 15:27:02 <mkolesni> manjeets, yes, also possibly other drivers will join as there has been interest from dragonflow, l2gw and even possibly midonet 15:27:35 <manjeets> neutron-lib is better atleast some of the maintenance will be offloaded from n-odl 15:27:44 <mkolesni> as to the location, the consensus reached with neutron PTL is to place it where is most convenient for us so it could be neutron or neutron-lib 15:28:11 <mkolesni> since theres a db migration involved this probably fits in neutron more 15:28:19 <mkolesni> but it can always move as necessary 15:28:37 <mkolesni> anyway back to the benefits 15:29:02 <manjeets> ok neutron and neutron-lib both have good attention from cores so maintenance will be good unless they spin a new repo for all this sharing stuff 15:29:17 <mkolesni> since the implemntation is simple and will be an infrastructure, our code will be much simpler so easier to maintain and less bug prone 15:29:35 <mkolesni> and also a major aspect i predict a benefit is performance 15:29:42 <mkolesni> here the benefit is twofold 15:30:13 <mkolesni> one is because the scalability of the n-ovn solution is much better since it scales per resource whereas journal scales per operation 15:30:21 <manjeets> mkolesni, how about CI (it will be reliable on outside repo, so can get little tricky sometimes) 15:30:31 <mkolesni> so it puts a very small overhead per resource 15:30:51 <mkolesni> manjeets, lets discuss this in a sec 15:31:10 <mkolesni> also the other benefit is that in the n-odl model the journal is a bottleneck 15:31:28 <mkolesni> so even for normal operation everything must go through the journal queue to get handled 15:32:12 <mkolesni> in the n-ovn solution only the erronous stuff will get handled in a queue, while normal operation will just happen as part of the normal synchronous flow of each operation 15:32:31 <mkolesni> i.e. if i create a port, the postcommit will actually try to sync the port to odl 15:32:43 <mkolesni> if it succeeds, great, were done with that 15:32:58 <mkolesni> and only if it fails will it be picked up later by the maintenance thread 15:33:13 <manjeets> mkolesni, so there will be journaling for failed operations only ? 15:33:16 <mpeterson> an added potential advantage we discussed today is that it might not even need locks :) 15:33:20 <mkolesni> this, i believe, with be more performant under load since you dont have a bottlenect 15:33:39 <mkolesni> mpeterson, indeed 15:34:10 <mkolesni> brb 15:34:44 <rajivk> mkolesni, thanks for listing, but i think, majority of the people don't know the approach. The better idea, will be to hand over these benefit to Peterson, he will compile them in good format and then everyone can relate all these things. 15:35:11 <mpeterson> rajivk: the spec lists the benefits and compares it to our journaling method 15:35:29 <rajivk> mpeterson, that is great. 15:35:36 <mpeterson> rajivk: and the new spec that I'm working on will probably expand on that 15:35:58 <mpeterson> manjeets: yes, failed operations will be defered to be handled by a maintenance task 15:36:08 <rajivk> mpeterson, if you have some data from PoC or networking-ovn data, it will be excellent. 15:36:11 <mkolesni> ok back 15:36:27 <mpeterson> rajivk: what kind of data ? 15:36:27 <mkolesni> im not sure what your concern about ci is, manjeets 15:36:44 <mkolesni> we will still have tests to see that odl is synced on various operations 15:36:45 <rajivk> anything which gives insight about scalability or performance improvement 15:37:23 <mpeterson> rajivk: I'll ask the ovn maintainer if he re ran the scalability/performance tests after implementing this solution 15:37:25 <mkolesni> mpeterson, you can ask lucas for the scale data, i know they did a scale run about a month ago 15:37:54 <mpeterson> #action mpeterson to ask OVN's maintainer for scale data on this solution, so it can be added to the spec 15:37:58 <manjeets> mkolesni, not really concerns, but it may get little hard to get fixes, it ci is broken because of changes in outside repo 15:38:03 <rajivk> mpeterson, this kind of data will make spec more impactfull 15:38:13 <mpeterson> rajivk: indeed 15:38:27 <mpeterson> manjeets: it's like when neutron makes a change, I also don't see the concern :/ 15:38:36 <mpeterson> manjeets: or any other dependency for that matter 15:39:13 <manjeets> yes, as I said not really a concern, we would be little more reliant on dependencies 15:39:15 <mkolesni> the ci for odl will remain a part of n-odl 15:39:33 <mpeterson> said that, like I mentioned to you today mkolesni, I think neutron-lib would be better suited as networking-* cores are also cores of neutron-lib 15:39:38 <mkolesni> just the ci for this infra will be there, but i expect it not to change that much anyway 15:39:59 <mkolesni> mpeterson, yes but afaik it doesnt have db migration scripts 15:40:10 <mkolesni> so how would you tackle the need? 15:40:40 <mpeterson> mkolesni: we can talk with the neutron core team and see what they say 15:41:07 <mpeterson> anyways, I do agree with mkolesni that the concerns over the CI shouldn't be too many 15:41:48 <mkolesni> mpeterson, you can write both options in the spec and see the replies 15:41:56 <manjeets> mkolesni, ++ 15:42:02 <mpeterson> indeed 15:42:22 <mkolesni> ok so mpeterson i working on the spec and should have a draft ready next week, right? 15:42:30 <mpeterson> mkolesni: correct 15:42:39 <mkolesni> cool 15:42:49 <manjeets> looking forward to review the spec 15:43:18 <mpeterson> oh, and btw, we want to target rocky 15:43:33 <mkolesni> the goal is to have it as poc in rocky 15:43:34 <mpeterson> so this work should be done before July 23rd 15:43:37 <manjeets> that might be tight though 15:43:45 <mkolesni> ans stabilize on stein 15:43:52 <mpeterson> right 15:44:04 <mkolesni> the code is already there but needs to be extracted 15:44:26 <mpeterson> indeed 15:44:28 <mpeterson> and it's quite simple 15:44:30 <mpeterson> let me show you 15:44:47 <mpeterson> https://github.com/openstack/networking-ovn/blob/f3f5257fc465bbf44d589cc16e9ef7781f6b5b1d/networking_ovn/db/revision.py 15:44:58 <mpeterson> 128 lines for the core code 15:45:19 <mpeterson> an extra 62 for the maintenance utilities https://github.com/openstack/networking-ovn/blob/f3f5257fc465bbf44d589cc16e9ef7781f6b5b1d/networking_ovn/db/maintenance.py 15:45:31 <mkolesni> btw on the ptg we had positive response from the neutron ptl and the people there, so its attainable i believe 15:46:03 <manjeets> question does ovn has db migration scripts inside it ? 15:46:11 <manjeets> n-ovn** 15:46:15 <mpeterson> yes, and a strong and well polished spec based on the one from OVN + OVN's success will allow it to move even faster 15:46:23 <mpeterson> manjeets: yes 15:47:16 <mpeterson> manjeets: you can find it in networking-ovn/db/ 15:48:11 <mpeterson> I think we won't have time to do the bug triage today 15:48:23 <mpeterson> so, any other questions regarding this? 15:48:28 <mpeterson> or we can adjourn the meeting? 15:48:33 <manjeets> next time then 15:48:54 <manjeets> I guess im good to go, need to read spec before I have questions 15:48:59 <mkolesni> i think everyone can read the n-ovn spec 15:49:11 <mpeterson> manjeets: offtopic, you are in India or USA? 15:49:15 <mkolesni> and then for next week when you have the spec ready we will go over that 15:49:28 <manjeets> manjeets, Oregon,Usa 15:49:31 <mpeterson> mkolesni: sounds good, I'll add it as an action 15:49:32 <manjeets> mpeterson, ^^ 15:49:50 <mpeterson> manjeets: nice :) 15:49:50 <mkolesni> oh no manjeets is talking to himself ;) 15:49:58 <mpeterson> hahaha it already started 15:50:13 <manjeets> mkolesni, m and tab magic 15:50:19 <mkolesni> indeed 15:50:24 <mkolesni> so many options :) 15:51:08 <mpeterson> #action all to read the OVN spec to familiarize, will be a topic next week's meeting for more input from everyone 15:51:20 <mpeterson> so, that's it... 15:51:20 <mkolesni> cool 15:51:22 <mpeterson> #topic Bug triage 15:51:29 <mpeterson> #info skipped, lack of time 15:51:39 <mpeterson> that's it 15:51:41 <mpeterson> #endmeeting