19:01:58 <oanson> #startmeeting Dragonflow
19:01:58 <openstack> Meeting started Mon Aug 14 19:01:58 2017 UTC and is due to finish in 60 minutes.  The chair is oanson. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:01:59 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:02:00 <dimak> hey
19:02:02 <openstack> The meeting name has been set to 'dragonflow'
19:02:03 <mlavalle> o/
19:02:08 <lihi> Hi
19:02:15 <leyal> Hi
19:02:24 <oanson> Hi everyone.
19:02:40 <oanson> Seeing as we started a couple of minutes late, I guess we can dive right in
19:02:48 <oanson> #info mlavalle dimak lihi leyal in meeting
19:02:56 <oanson> #topic Roadmap
19:03:26 <oanson> LBaaS - Finally managed to make some headway. I have updated the spec. I think it's ready for initial reviews
19:03:51 <oanson> #link LBaaS spec: https://review.openstack.org/#/c/477463/
19:04:01 <dimak> I'll post my comments tomorrow, had to do some extra background reading
19:04:21 <lihi> Nice. I'll review it
19:04:22 <oanson> Yes. The v2 API is not trivial.
19:05:08 <oanson> There are a couple of interesting points inside that affect the entire project
19:05:32 <leyal> Yep , specially  the separation between load-balancer  and listener ..
19:05:33 <oanson> e.g. model polymorphism, and greater than OpenFlow functionality
19:05:52 <oanson> Yes.
19:06:01 <dimak> oanson, a question
19:06:05 <oanson> Shoot
19:06:32 <dimak> The API we implement there will get changed to support common classification framework?
19:06:43 <oanson> I don't know
19:06:43 <dimak> mlavalle, tagging you as well
19:07:00 <mlavalle> don't know either
19:07:17 <oanson> But we should prepare for that. Or we could consider doing that ourselves, and converting LBaaS listener API to common classification
19:07:21 <mlavalle> is this implemenation supposed to implement the upstream LBaaS v2 api?
19:07:27 <oanson> Yes
19:07:29 <mlavalle> ok
19:07:42 <mlavalle> I'll also review the patchset with that in kind
19:07:45 <mlavalle> mind^^^^
19:07:51 <oanson> That would be great!
19:08:14 <mlavalle> it's not my area of expertise, but might as well get familiar with it
19:09:21 <oanson> I tried to be as clear as possible. I would like this spec to be clear to people who don't know Neutron's LBaaS v2 inside and out.
19:09:35 <mlavalle> ok, great
19:09:55 <oanson> I have found very little information aside from the Neutron documentation, so once the spec stabilises, I might do well to also extract a blog post.
19:10:16 <oanson> Or improve my googling skills. Either way :)
19:10:21 <oanson> Anything else for LBaaS?
19:10:39 <leyal> there is some documentation in the octavia  project
19:10:48 <leyal> not to much ..
19:11:17 <oanson> All right.
19:11:36 <oanson> RPM packaging - I haven't made any progress. But I'll try to upload a draft by the end of the week.
19:12:27 <oanson> L3 flavour - dimak, any progress?
19:12:44 <dimak> No, I haven't made much progress
19:13:06 <mlavalle> dimak: you've been working on the dragonflow side of this, right?
19:13:25 <dimak> mlavalle, what do you mean?
19:13:57 <dimak> I have a spec up and some draft of l3 service provider for dragonflow l3
19:13:57 <johnsom> Yes, as we move towards deprecating neutron-lbaas, all of the documentation is moving here: https://docs.openstack.org/octavia/latest/
19:13:59 <mlavalle> dimak: we need to implement something in the L3 side of upstream, right?
19:14:54 <dimak> mlavalle, my plan is l3 service provider + a dragonflow l3 agent
19:15:04 <oanson> johnsom, thanks. I'll take a look.
19:15:23 <mlavalle> dimak: correct. you addressing both things?
19:15:30 <dimak> yes
19:15:41 <mlavalle> great
19:16:02 <mlavalle> just trying to see if help is needed
19:16:19 <mlavalle> at the very least, I'll review the patchsets
19:16:35 <oanson> Great. Thanks!
19:16:50 <oanson> Anything else for L3 flavour?
19:16:52 <dimak> mlavalle, I haven't made a lot of headway there, but I'll let you know when its up
19:17:00 <mlavalle> :-)
19:17:25 <oanson> All right. lihi, what about etcd publisher?
19:17:42 <oanson> #link etcd3 driver https://review.openstack.org/#/c/489246/
19:18:03 <oanson> lihi, the etcd3 patch above is finished, right?
19:18:30 <lihi> Yes. Just minor fixes are missing.
19:18:45 <lihi> I'll remove the WIP tag tomorrow.
19:18:49 <oanson> Cool.
19:18:59 <oanson> What about the publisher itself?
19:19:15 <lihi> The pubsub is coming along nicely. I'll upload it in the following days
19:19:37 <oanson> Great.
19:20:06 <oanson> Do you plan on adding the OSA gate to dragonflow soon? Or do you want to give it a sanity test first?
19:21:28 <lihi> I'll start to do it next week, I think. I want to make sure before that the pubsub is working nicely
19:21:54 <oanson> Sure.
19:22:12 <oanson> All right. That's about it for roadmap.
19:22:17 <oanson> Anything else before we move on?
19:23:24 <oanson> One last thing - I think we'll start looking at new features only after the tag date. Let's concentrate on bugs and stability for the next few weeks.
19:23:31 <oanson> #topic Bugs
19:24:00 <dimak> +1
19:24:18 <dimak> Tag date would be last date of august?
19:24:31 <oanson> Let me check
19:24:42 <lihi> Any major bugs blocking us?
19:24:48 <oanson> 1st of September.
19:25:00 <dimak> lihi, blocking us from what? :P
19:25:18 <oanson> Bug 1708178 is blocking Kuryr integration. leyal, I understand you are working with irenab on this?
19:25:19 <openstack> bug 1708178 in DragonFlow "LBaaSv2 with 3rd party provider does not work if L3agent is disabled" [Critical,New] https://launchpad.net/bugs/1708178
19:25:36 <oanson> (I think everyone is working on this, but I think you are the main assignee at the moment)
19:26:26 <leyal> yes , currently i can't reproduce it - just by creating a LB ..
19:26:51 <oanson> Have you tried to go through the Kuryr flow? Or do you need irenab for that?
19:27:09 <leyal> i need irenab for that ..
19:27:44 <dimak> I've resurrected https://bugs.launchpad.net/dragonflow/+bug/1686108
19:27:45 <openstack> Launchpad bug 1686108 in DragonFlow "DHCP port does not have an ARP responder" [Medium,New]
19:27:47 <leyal> We should try reproduce it tomorrow  ..
19:27:56 <oanson> leyal, Sounds like a plan.
19:27:57 <dimak> Changed it to include ICMP as well
19:28:06 <oanson> dimak, still Medium?
19:28:28 <dimak> Well, its blocking tempest
19:28:32 <dimak> but DHCP works
19:28:33 <dimak> so..
19:28:58 <oanson> If it's blocking tempest, it's high. Because tempest is high
19:29:09 <dimak> Updated
19:29:44 <oanson> Thanks.
19:29:53 <dimak> I've uploaded a new version to fix https://bugs.launchpad.net/dragonflow/+bug/1636829
19:29:54 <openstack> Launchpad bug 1636829 in DragonFlow "Conflict between flat network and DNAT app" [High,In progress] - Assigned to Dima Kuznetsov (dimakuz)
19:30:16 <leyal> dimak ,  i started to work on that .. i can upload a patch - but still need to write full-stack test ..
19:30:30 <dimak> I don't think it passes all fullstack tests, and it sits on quite a few patches, but hopefully they'll go in this time
19:30:42 <dimak> leyal, the ICMP thing?
19:30:48 <oanson> I see we have other dhcp bugs. 6-7 small stuff. leyal, when you get back to hacking the DHCP app (after the Kuryr stuff), you might want to skim through to see what's still relevant, and what can be easily closed.
19:31:07 <dimak> leyal, we have tempest failing on that, if tempest is happy, the fix is good
19:31:31 <mlavalle> LOL, let's strive to make Tempest happy
19:31:34 <leyal> dimak , sure - i will upload a patch a check ..
19:31:49 <oanson> Great!
19:31:50 <dimak> leyal, cool
19:31:50 <leyal> oanson, sure ..
19:32:06 <oanson> dimak, regarding bug 1636829 - I see you bumped it down to high
19:32:07 <openstack> bug 1636829 in DragonFlow "Conflict between flat network and DNAT app" [High,In progress] https://launchpad.net/bugs/1636829 - Assigned to Dima Kuznetsov (dimakuz)
19:32:08 <dimak> Send me the link, I'll add a Depends-On on my tempest patch
19:32:18 <dimak> It was critical before
19:32:34 <oanson> Yes. Why isn't it critical any more? Is DNAT working out of the box?
19:32:35 <dimak> But DNAT works as long as you don't have L3 agent
19:33:03 <apuimedo> my ears are ringing
19:33:12 <apuimedo> did somebody say the work Kuryr?
19:33:30 <oanson> apuimedo, many times :)
19:33:38 <apuimedo> :-)
19:34:09 <oanson> dimak, all right. I trust your judgement here. :)
19:34:09 <apuimedo> oanson: I just managed to get kuryr-kubernetes to create pods and services on ipv6 only
19:34:11 <apuimedo> :-)
19:34:28 <oanson> Cool!
19:34:31 <dimak> 👏👏👏
19:35:03 <lihi> 🎉🎉🎉
19:35:35 <oanson> If it's all the same with everyone, I'll stick to 7-bit single-byte characters :D
19:36:03 <dimak> so 90s
19:36:20 <oanson> apuimedo, we're working hard to get the kuryr-dragonflow integration working. We're hoping to have a gate job for that too.
19:37:11 <oanson> dimak, the port locality stuff - that's blocking bug 1705503, right?
19:37:12 <openstack> bug 1705503 in DragonFlow "Trunk subport will not be available after controller restart" [High,Fix committed] https://launchpad.net/bugs/1705503
19:37:32 <dimak> Not sure anymore
19:37:54 <dimak> But it should work with https://review.openstack.org/#/c/493282/
19:37:58 <oanson> Could you elaborate?
19:38:20 <apuimedo> oanson: a gate would be awesome
19:38:32 <apuimedo> oanson: so did you fix the trunk active status thing?
19:38:41 <dimak> We've changed some port logic since I reported this, I have to check how it all behaves now
19:39:28 <oanson> All right. Please update the port locality bug with any new information
19:39:33 <dimak> oanson, but at the moment, restart will cause other troubles, the patch that adds port status removed binding so currently controller will throw the events away
19:39:46 <dimak> oanson, good thing that patch is just 5 commits away from master :P
19:40:12 <oanson> Yeah... But I have issues with the naming of the patch below it :)
19:40:17 <oanson> apuimedo, yes. We have other bugs now, but that bug is solved :)
19:40:33 <oanson> dimak, do you need another pair of hands?
19:40:47 <apuimedo> oanson: we'll be able to eat bugs once meat runs out. Software will provide
19:41:21 <dimak> oanson, I think I'm on top of it, the whole chain passes CI :)
19:41:31 <oanson> apuimedo, LOL
19:41:45 <oanson> dimak, glad to hear it.
19:42:02 <dimak> apuimedo, I'll have only range-free software bugs
19:42:17 <oanson> dimak, does that include off-by-one?
19:42:35 <oanson> All right. There are some medium bugs that I think we can close already, but they can wait (by definition).
19:42:51 <oanson> Anything else for bugs? Recipes, maybe?
19:43:26 <oanson> #topic Open Discussion
19:43:31 <oanson> Fire away!
19:43:40 <mlavalle> I have two things
19:44:09 <oanson> Shoot.
19:44:19 <mlavalle> 1) Next Monday is my first day at Huawei. So next meeting I'll see you at the other side
19:44:40 <mlavalle> 2) What is the DB we prefer with dragonflow?
19:44:50 <oanson> 1) Cool. Welcome aboard!
19:45:03 <dimak> 🎉🎉🎉
19:45:06 <oanson> 2) Redis, or etcd+zmq - since these are the ones we gate.
19:45:16 <apuimedo> dim mine are usually out of bounds, so I lose them
19:45:32 <mlavalle> ok, I'm working on the Vagrant in the repo
19:45:40 <mlavalle> I'll try to make it work
19:45:43 <oanson> Once lihi finishes the etcd3 work, we'll probably gate something else alongside zmq to test the 'external publisher' capability.
19:46:07 <dimak> oanson promised to write a blog post on his vagrant scripts
19:46:13 <lihi> IMHO Redis is more stable
19:46:24 <mlavalle> yeah, but we already have something in the repo
19:46:26 <oanson> I did...
19:46:29 <lihi> For now, at least
19:46:52 <mlavalle> to promote adoption, I think that what is in the repo should work
19:47:05 <oanson> That's definitely true.
19:47:54 <oanson> But that just means we can also remove them and link to the openstack-vagrant project. It is tested extensively, since all of us use it.
19:48:08 <oanson> We definitely can't have a vagrant example that doesn't work.
19:48:51 <mlavalle> that's all I had
19:49:14 <oanson> mlavalle, I think any issues you'll run into are devstack configuration issues. Do you think you could try it out, and identify issues and report them?
19:49:29 <mlavalle> oanson: definitely
19:49:33 <oanson> I think this, since we deploy dragonflow constantly,
19:49:35 <oanson> Thanks!
19:49:51 <mlavalle> which vagrant are we talking about. the one in the repo?
19:49:58 <oanson> Yes
19:50:04 <mlavalle> will do
19:50:06 <mlavalle> :-)
19:50:38 <oanson> Unless we decide to move completely to openstack-vagrant. But I'm not sure it's mature enough, and it doesn't have an auto-multi-host-deployment feature yet.
19:51:07 <oanson> If anyone has any thoughts on this, now would be good :)
19:52:01 <oanson> All right. So let's stick with the current plan. Yes, the vagrant in repo :)
19:52:11 <mlavalle> \o/
19:52:17 <oanson> Anyone else wants a go?
19:53:01 <oanson> Great.
19:53:05 <oanson> Thanks everyone for coming!
19:53:09 <mlavalle> o/
19:53:14 <dimak> Bye
19:53:28 <lihi> Bye
19:53:30 <oanson> #endmeeting