08:00:05 #startmeeting Dragonflow 08:00:06 Meeting started Mon Feb 5 08:00:05 2018 UTC and is due to finish in 60 minutes. The chair is oanson. Information about MeetBot at http://wiki.debian.org/MeetBot. 08:00:07 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 08:00:09 The meeting name has been set to 'dragonflow' 08:00:15 hi 08:00:23 Yo 08:00:26 Hi 08:00:31 Who's here for the weekly? 08:00:38 irenab, you joining us? 08:00:46 dimak is in fosdem, so I doubt he'll be here 08:00:46 hi 08:01:01 lihi, you in? 08:01:10 Hi 08:01:21 All right! Let's get this party started! 08:01:25 #topic Roadmap 08:01:32 DNS - lihi? 08:01:57 No update. Was working mostly on Deployment 08:02:07 Sure 08:02:13 Upgrades - 08:02:24 (One sec, I got logged out for some reason) 08:02:47 No real update. Sorry. 08:02:55 Deployment - 08:03:00 OSA - lihi ? 08:03:36 I still don't have a working env, but I think it's coming along. I'm still trying to get dragonflow to communicate with the DB 08:03:47 Sure. 08:04:04 lihi: if you want me to help, just say so 08:04:19 ok, I will. Thanks 08:04:21 RPM packaging - patch currently breaks due to release notes in docs in pike 08:04:40 The patch https://review.openstack.org/#/c/539472/ should solve it 08:04:44 But waiting for reviews :( 08:05:02 Import Neutron DB - CrayZee ? 08:05:22 No progress there. I am currently working on SkyDive 08:05:27 Sure 08:05:35 Gates - Grenade gate 08:05:56 Added in patch https://review.openstack.org/#/c/496837/ , but currently failing miserably. 08:06:06 Will probably reduce tempest there to a bare minimum just to get it going 08:06:35 Tempest gate - Uploaded a first patch in trying to resolve bug #1740739 08:06:36 bug 1740739 in DragonFlow "Security group mismatch for floating IP" [High,New] https://launchpad.net/bugs/1740739 08:06:36 oanson, ay knows issues? 08:06:50 for the grenade 08:06:58 irenab, it fails on the initial tempest test 08:07:18 grenade works by installing initial version, running tempest tests for sanity, upgrading, and then running the tempest tests again 08:07:32 Since we fail the tempest tests the first time around, the gate refuses to cooperate 08:08:01 so in general tempest fails, regardless of grenade? 08:08:14 irenab, yes. That's bug #1740739 08:08:16 bug 1740739 in DragonFlow "Security group mismatch for floating IP" [High,New] https://launchpad.net/bugs/1740739 08:08:24 That's the last hurdle 08:08:45 cannot this test be temporary skipped? 08:08:56 It can. 08:09:00 oanson: I see that https://review.openstack.org/#/c/539546/ still fails on tempest on cross-tenant-traffic 08:09:18 CrayZee, yes. It is only a first patch in a set. 08:09:27 ok 08:09:35 SecGroups wasn't updated to take reg6 into account instead of source IPs. 08:09:47 Since that app is a bit of a mess, it's taking me longer than I thought it would :) 08:09:58 And there will probably be other underlying problems. 08:10:24 Especially for multi-node stuff, which sadly isn't even gated. 08:10:31 irenab, you said you had an update on this ^^^ ? 08:10:50 oanson, I had a chat with dmellado last week, he may have something to start with 08:11:07 That would be great! 08:11:16 but it was before zuulv3 move 08:11:30 so will need some modifications 08:11:48 irenab, something major? 08:11:55 irenab: I can do them, as leyal and I handled most of the zuul3 move 08:12:21 oanson, I do not know. CrayZee , would be great, lets get in touch with dmellado later today 08:12:49 irenab: I prefer tomorrow, when I will be at the office :) 08:12:54 sure 08:12:59 Sounds like a plan! 08:13:10 I guess this also covers the kuryr integration gate 08:13:21 Troubleshooting - 08:13:28 CrayZee, you mentioned skydive? :) 08:14:10 yes, I have encountered a few issues due to the fact we need the eventlet monkey_patch 08:14:36 then we decided to go in the direction of having it as a separate daemon 08:15:06 there are issues, though with that approach, as we will not get any notification for added/removed items 08:15:33 the only solution I can see at the moment is to delete all the items and re-create everything on every run 08:15:44 this is not really a good solution 08:15:55 Actually, as a separate daemon, we are not limited with the events the DF controller gives us 08:16:20 We can subscribe to the pub/sub mechanism and handle the events manually in a way fitting the skydive daemon 08:16:20 Lihi proposed using the pubsub 08:16:26 CrayZee, what should skydive agent do? 08:16:47 The only issue with a separate daemon is that we don't know what the DF controller 'sees', i.e. what's in its db_store. 08:16:57 irenab: visualize the topology view, the way DF sees it 08:17:06 But I guess this can be worked around... 08:17:21 oanson: that would require developing yet another subscriber that is only for skydive 08:17:27 CrayZee, limited to local controller view or system wide? 08:18:02 irenab, I also have a hope it will be an extended platform for applications to visualize their view on the system. e.g. security groups shows which ports can communicate, and trunk/nested-ports show which port is behind which 08:18:22 CrayZee, oanson I think this feature may require some level spec 08:18:42 CrayZee, why can't we reuse the code in DF controller? (And if there is a reason, let's fix that) 08:18:49 oanson, still not sure if its local view or system view 08:18:53 or both 08:18:59 irenab: depending on where it runs and who updates it. As I see it, there is one analyser (the visualizer) and every node will send its information to it. the analyser aggregates it 08:19:27 oanson: etcd subscriber for starter 08:19:37 I think irenab is right. There may be some confusion here. Since each controller tags the objects it sees with its chassis name 08:19:48 CrayZee, there is a config and runtime data that may differ 08:20:01 CrayZee, not sure I understand. Why can't the etcd subscriber be used in the skydive daemon? 08:20:22 oanson: as it required eventlet, and we are on square 1 08:20:58 Oh for crying out loud :( 08:21:14 CrayZee, oanson seems there is a mizture of scope and implementation questions 08:21:16 irenab: I can show you a little demo tomorrow, and align on that... 08:21:27 We can write another etcd subscriber that use a different etcd-python library :) 08:21:38 lihi, a question re the etcd3gw package - is it api compatible with the non-eventlet version? 08:21:58 CrayZee, this can be great for a start, but I think it may still need some spec/devref 08:22:24 irenab: thought so too (though I hate writing them :) ) 08:22:45 I don't know. This needs to be checked. It was written specifically to be used with eventlet 08:23:07 CrayZee, are there other issues with eventlet except the trollius/select issue? 08:23:25 (I feel like trollius is trolling us) 08:23:56 oanson: I do not remember, but my guts say there is... :P 08:24:31 All right. Let me know tomorrow when you had another look at it. A separate daemon may be a good idea regardless of the eventlet issue 08:24:37 Let's take the rest offline. 08:24:45 Agreed? 08:24:50 yup 08:25:19 Headless Dragon (Dragonflow+Kuryr+k8s) - leyal ? 08:26:04 oanson, Headless Dragon sounds like poor man solution :-) 08:26:25 That's what Xykon thought, too... 08:27:44 leyal, you in? 08:27:47 don't done much progress with that 08:28:18 No worries. 08:28:22 focused mostly on kuryr tasks last week.. 08:28:29 Sure 08:28:34 Anything else for roadmap? 08:29:23 #topic Bugs 08:29:26 Anything for bugs? 08:29:58 #topic Open Discussion 08:30:01 Anything we left out? 08:31:05 All right 08:31:09 Thanks for coming 08:31:14 #endmeeting