08:00:20 <oanson> #startmeeting Dragonflow 08:00:21 <openstack> Meeting started Mon Aug 21 08:00:20 2017 UTC and is due to finish in 60 minutes. The chair is oanson. Information about MeetBot at http://wiki.debian.org/MeetBot. 08:00:22 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 08:00:24 <openstack> The meeting name has been set to 'dragonflow' 08:00:30 <dimak> Hey 08:00:42 <kkxue> Hi,guys 08:01:11 <oanson> Hi all. 08:01:23 <oanson> Let's wait another minute. Maybe lihi and irenab will join us. And then we'll start 08:01:41 <lihi> Hi 08:01:52 <itamaro> hi everyone 08:02:35 <oanson> All right! Let's start 08:02:43 <oanson> #info dimak kkxue lihi itamaro in meeting 08:02:47 <oanson> #topic Roadmap 08:02:53 <irenab> hi 08:03:11 <oanson> Before we start with actual roadmap items, I want to post the trello board for Dragonflow 08:03:19 <oanson> #link Dragonflow Trello board https://trello.com/b/PM7nah4Z/project-dragonflow 08:03:33 <oanson> I am not sure how well using Trello will work out, but we'll see :) 08:03:52 <oanson> #info irenab also in meeting 08:04:09 <oanson> All right. Roadmap: The LBaaS spec is uploaded 08:04:15 <oanson> I think it's ready for a real review 08:04:24 <oanson> #link LBaaS spec https://review.openstack.org/#/c/477463/ 08:05:07 <irenab> oanson, added to my todo list 08:05:07 <oanson> There's no real rush with it, as it is physically impossible to get it into pike 08:06:04 <oanson> Once we went over it, it might be helpful to ask the Kuryr and Octavia people to look at it as well. 08:06:32 <apuimedo> oanson: look at the lbaas spec? 08:06:39 <oanson> apuimedo, yes 08:06:43 <apuimedo> ok 08:06:52 <oanson> I'd be happy to hear feedback. 08:06:55 <irenab> oanson, I can assure you, we will test it too 08:07:20 <oanson> irenab, I didn't doubt it for a second :) 08:07:37 <oanson> Anything else for LBaaS? 08:07:52 <irenab> oanson, just one comment 08:07:58 <oanson> Shoot 08:08:11 <irenab> we should keep the option to integrate 3rd party LB with dragonflow 08:08:27 <oanson> Yes, we should 08:08:58 <oanson> irenab, could you leave a comment on the spec so it won't be forgotten? 08:09:06 <irenab> sure 08:09:10 <oanson> Thanks. 08:10:20 <oanson> L3 flavour - dimak, any progress? 08:10:25 <oanson> Or were you bug-hunting? 08:10:52 <dimak> oanson, not this week, was working on other stuff this week 08:11:02 <oanson> Sure 08:11:21 <oanson> ETCD publisher - lihi, I saw you uploaded a WIP for the publisher 08:11:23 <dimak> I'll focus on it after we tag pike 08:11:31 <oanson> And updated the etcd3 patch to be non-wip 08:11:43 <oanson> dimak, sure. That's the timeframe we discussed 08:12:15 <oanson> #link etcd3 driver https://review.openstack.org/#/c/489246/ 08:12:20 <lihi> Yeah. Right now I'm working on getting it installed with devstack :) 08:12:43 <oanson> Cool 08:13:07 <oanson> Once that's done the next step is the OSA deployment with it? 08:13:31 <lihi> And make sure it really works 08:14:08 <oanson> Very cool! 08:14:51 <oanson> I didn't get to work on RPM packaging this week. I won't make any promises this time, since there are a couple of patches I feel need to get in before tag date since they change DB structure 08:15:00 <oanson> Sorry, anything else for etcd? 08:15:16 <lihi> No 08:15:32 <oanson> All right. Anything else for roadmap? 08:16:24 <oanson> #topic Bugs 08:16:46 <oanson> Let's start with bug 1708178 08:16:47 <openstack> bug 1708178 in DragonFlow "LBaaSv2 with 3rd party provider does not work if L3agent is disabled" [Critical,New] https://launchpad.net/bugs/1708178 08:17:01 <oanson> I understand it can be worked around with selective-proactive disabled? 08:17:06 <dimak> I suggest to close it as invalid 08:17:20 <dimak> It is not related to selective proactive IMO 08:17:32 <dimak> its just that selective proactive is rather broken 08:17:35 <irenab> oanson, checking it now for BM, and we saw it works for nested environment 08:17:39 <oanson> I understood it works without selective-proactive, but not with it 08:18:00 <oanson> irenab, but with selective-proactive off? 08:18:04 <dimak> same way routing through shared router doesn't work 08:18:09 <irenab> oanson, yes 08:18:21 <dimak> lbaas integration itself works alright 08:18:46 <oanson> There are two options. 08:18:53 <oanson> We either update the bug, or open a new one instead. 08:19:05 <dimak> It's a regular 'VM' port as far as dragonflow is concerned 08:19:14 <dimak> I suggest a new one 08:19:15 <oanson> Selective-proactive being broken this way is still critical, since we claim that it's working 08:19:21 <dimak> I might have one for it 08:19:29 <oanson> dimak, linky? 08:19:34 <dimak> looking 08:19:48 <irenab> oanson, shall we set deafault to selective proactive disabled till it fixed? 08:19:55 <oanson> Also, how complicated would it be to see if https://review.openstack.org/#/c/480196 solves the issue? 08:20:23 <oanson> irenab, if we have to, but let's wait till the tag cut-off for it. I don't think a week will make much of a difference in this case. 08:20:46 <oanson> Scratch that - we'll disable it now, and re-enable it when the bug is fixed 08:21:05 <irenab> +1 08:21:16 <dimak> oanson, I'll find it after the meeting, if there isn't one I'll open a new bug 08:21:25 <oanson> dimak, sure. 08:21:47 <oanson> I am still interested to know if it's possible to test again with https://review.openstack.org/#/c/480196 applied? 08:22:01 <oanson> (Including all dependencies) 08:22:07 <dimak> We can test it in tempest ;) 08:22:15 <oanson> We should 08:23:14 <oanson> dimak, irenab ? 08:23:29 <dimak> Sure, I can try bringing up an env 08:23:51 <oanson> Thanks. If you prefer, you can also walk me through the process. 08:24:33 <oanson> dimak, note that this bug also has l3-agent support elements. Might be worth checking and updating in subsequent related bugs as well 08:24:47 <oanson> Bug 1658479 08:24:48 <openstack> bug 1658479 in DragonFlow "tenant_id is still being used" [High,In progress] https://launchpad.net/bugs/1658479 - Assigned to Omer Anson (omer-anson) 08:25:09 <oanson> I uploaded a WIP for it, https://review.openstack.org/#/c/480196 08:25:09 <dimak> oanson, I skimmed over the patch, does it subscribe to new topics? 08:25:29 <oanson> It should be an in-place replacement, so it should 08:26:00 <oanson> The first question is do we want it at this time? According to https://specs.openstack.org/openstack/neutron-specs/specs/newton/moving-to-keystone-v3.html#neutron-api , tenant_id isn't removed yet 08:26:08 <oanson> And it will stay for a while. 08:26:34 <oanson> And the patch forces us to support both, which is... wrong... 08:26:52 <dimak> then lets wait with all the models that are not through yet 08:27:29 <oanson> Right. So I'm bumping the bug back to medium. I'll leave the patch for later. 08:27:51 <oanson> Bug 1684108 08:27:53 <openstack> bug 1703334 in pulseaudio (Ubuntu) "duplicate for #1684108 start-pulseaudio-x11 tries to load /usr/share/sounds/ubuntu/stereo/bell.ogg which is missing in Kubuntu" [High,Fix released] https://launchpad.net/bugs/1703334 - Assigned to Daniel van Vugt (vanvugt) 08:28:12 <dimak> Heh, that's not the same one, openstack 08:28:20 <oanson> Let's try again 08:28:21 <oanson> Bug 1684108 08:28:37 <oanson> DHCP port does not have an ARP/ICMP responders 08:28:41 <oanson> https://bugs.launchpad.net/dragonflow/+bug/1686108 08:28:42 <openstack> Launchpad bug 1686108 in DragonFlow "DHCP port does not have an ARP/ICMP responders" [High,In progress] - Assigned to Eyal Leshem (leyal) 08:28:55 <oanson> Should be solved by either https://review.openstack.org/#/c/493794/ or https://review.openstack.org/#/c/495544/ 08:29:13 <dimak> Yes, I opted for a quick fix 08:29:46 <dimak> I think we should remove it and move it into 'install subnet dhcp flows' once we refactor dhcp app a bit 08:30:04 <oanson> dimak, I don't understand, sorry. 08:30:49 <dimak> I say that we should {,un}install responders at the same time we install the rest of subnet dhcp related flows 08:31:10 <oanson> So the way it's done in leyal's patch 08:31:12 <dimak> yes 08:31:25 <dimak> but to get it before pike, I proposed to do it separately 08:31:48 <oanson> Then let's go with that. You think you can update leyal's patch to pass review? (leyal is out for the week, and the tag cut-off is coming) 08:32:04 <oanson> Or do you want me to do it? 08:32:10 <oanson> Or do we think that it can wait a week? 08:32:19 <dimak> what do refer to as that 08:32:39 <oanson> dimak, sorry? 08:32:55 <dimak> >> Then let's go with that 08:33:24 <oanson> Sounds like a plan. dimak, since I am also making changes in that area, let's coordinate offline 08:33:42 <dimak> ok 08:33:59 <oanson> Bug 1705503 08:34:00 <openstack> bug 1705503 in DragonFlow "Trunk subport will not be available after controller restart" [High,Fix committed] https://launchpad.net/bugs/1705503 08:34:12 <dimak> https://review.openstack.org/#/c/493282/ 08:34:18 <dimak> Deals with it 08:34:21 <dimak> reviews welcome 08:34:47 <oanson> Very cool. Reviewers, start picking it apart :) 08:35:12 <lihi> 👍 08:35:12 <dimak> oanson, any idea if there are trunk tests for tempest? 08:35:38 <oanson> dimak, no idea. 08:35:51 <oanson> But we'll have a kuryr-integration gate which will test this feature thoroughly 08:36:04 <dimak> I'll take a look later on 08:36:20 <oanson> Sure. Thanks 08:36:23 <oanson> Bug 1707496 08:36:25 <openstack> bug 1707496 in DragonFlow "Stabilize tempest job" [High,In progress] https://launchpad.net/bugs/1707496 - Assigned to Dima Kuznetsov (dimakuz) 08:36:38 <oanson> https://review.openstack.org/#/c/489003/ shows that only bgp is failing 08:37:00 <oanson> It works on my environment. I think the gate environment is missing some config. I'll try to update today. 08:37:28 <dimak> Hangs on 3 patches, https://review.openstack.org/#/c/493359/ https://review.openstack.org/#/c/483385/ and either of the DHCP responder patches 08:37:36 <irenab> dimak, https://github.com/openstack/neutron/blob/master/neutron/tests/tempest/scenario/test_trunk.py 08:37:45 <irenab> on your previous question 08:37:53 <dimak> irenab, thanks! 08:38:09 <dimak> I'll make sure we run it in our tempest job 08:38:16 <irenab> great 08:38:19 <oanson> Cool! 08:38:34 <oanson> Last High bug: 1708157 08:38:35 <openstack> bug 1708157 in DragonFlow "Implement port locality spec" [High,In progress] https://launchpad.net/bugs/1708157 - Assigned to Dima Kuznetsov (dimakuz) 08:38:54 <dimak> Yes 08:39:06 <dimak> Some work was done last week introducing bind events 08:39:11 <oanson> Is it blocking anything we need for pike? 08:39:16 <dimak> now we need to add distributed ports 08:39:19 <dimak> nope 08:39:37 <oanson> Then I want to bump it down to medium, and get back to it after that tag cut-off 08:39:56 <dimak> well, we could have solved DHCP responders with it I think, but we'll get to it later 08:40:37 <oanson> I don't think it can be finished *and* used before the tag cut-off. I don't want to tie things we can finish in advance on something that will take us more time 08:40:44 <dimak> Yes 08:40:52 <oanson> Yes it can, or yes it can't? 08:41:29 <dimak> Yes we don't need it before the tag 08:41:41 <dimak> easier to accomplish our goals without it for now 08:41:42 <oanson> So it's bumped down, unless there are objections? 08:42:43 <oanson> This was the last high bug for now. Unless we find something new, we're in good shape 08:43:18 <oanson> All right, anything else for bugs? 08:44:05 <oanson> #topic Open Discussion 08:44:10 <oanson> Fire away 08:45:16 <oanson> If that's the case, we can finish early :) 08:45:23 <oanson> Anything before we close? 08:45:36 <itamaro> Well, yes 08:45:49 <oanson> itamaro, the fire away :) 08:45:52 <oanson> then* 08:46:08 <itamaro> I want to bring up again the issue of flow tagging using coockie 08:46:22 <oanson> itamaro, linky? 08:46:37 <itamaro> one sec 08:47:11 <itamaro> > https://review.openstack.org/#/c/424123/ 08:47:56 <itamaro> it is basically a debug tool which allows each application to color it's flows. 08:48:29 <itamaro> the coloring itself, is defined in run time. 08:49:14 <itamaro> I was asked at the time to make fullstack tests for this feature. 08:49:45 <oanson> itamaro, let's start by rebasing the patch 08:49:51 <irenab> if this is a debug tool, do we need full stack tests for it? 08:50:00 <oanson> Let me also find something in the tests code 08:50:35 <itamaro> Ok. I'll start by doing a rebase. 08:51:04 <itamaro> sound like a plan. 08:51:09 <itamaro> :) 08:51:18 <oanson> We also have a test_app_base.DFAppTestBase base class that allows you to start a controller with a given list of apps 08:51:48 <itamaro> So only one app can do? 08:51:50 <oanson> A unit test can then read verify the correct apps write the correct cookie colour 08:52:04 <oanson> It's a list. I think several apps can be given. 08:52:38 <oanson> But maybe just a simple unit test to make sure the global cookie is registered will do. It's a unit test, not integration test, for a development feature 08:52:46 <oanson> That relies on a tested library 08:53:10 <itamaro> Yes I thing unit test will do. 08:53:16 <irenab> +1 08:53:25 <oanson> Great! 08:53:29 <oanson> Anything else? 08:54:54 <oanson> All right. Thanks everyone for coming 08:55:15 <oanson> #endmeeting