14:00:42 <dmellado> #startmeeting kuryr
14:00:45 <openstack> Meeting started Mon Apr  1 14:00:42 2019 UTC and is due to finish in 60 minutes.  The chair is dmellado. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:48 <openstack> The meeting name has been set to 'kuryr'
14:00:53 <dmellado> Hi folks, who's here today for the meeting?
14:02:30 <maysams> o/
14:02:35 <aperevalov> o/
14:02:39 <aperevalov> hi
14:03:38 <dmellado> hi! so, first of all, I've tagged kuryr-tempest-plugin for Stein
14:04:06 <dmellado> as well as created stable/stein branch, for which I've set 1.0.0 release after speaking with the foundation so we can follow the default release cadence now
14:04:17 <dmellado> congrats to all to finally arrive to 1.0.0 release
14:04:19 <dmellado> \o/
14:05:36 <dmellado> then, second, the CI has been broken since today, although some folks have been expressing issues since Thu
14:05:59 <dmellado> since that it could be related to the k8s but so far tempest jobs are failing due to some issues on reaching k8s api
14:06:21 <dmellado> so be aware that your changes will probably fail for a while until we get to fix that
14:06:55 <dmellado> any topic from your side maysams aperevalov?
14:08:48 <aperevalov> yes, one week ago I mention issue when we chaning to DPDK driver (binding problem). So it could be solved just by receiving pod resources. I wrote to openstack-kuryr, so dulek knows ;)
14:08:49 <maysams> dmellado: I'm trying to work with dulek to find out what is problem with the test that is not passing the upstream k8s tests. If the problem is our implementation or the test is wrong
14:09:14 <dmellado> maysams: yeah, IIRC you told me that that specific test is even skipped on their side, isn't it?
14:09:34 <maysams> dmellado: Yes, it's
14:09:58 <dmellado> it'd be interesting to get to know why it has been set to skipped
14:10:08 <aperevalov> also do your have now issues with performance when creating/binding ports?
14:10:13 <dmellado> maybe there's a valid reason for that and we should just do the same
14:10:23 <dmellado> aperevalov: what kind of issues?
14:11:21 <aperevalov> performance issues - I mean delay at port creation time or in the binding driver
14:13:14 <dmellado> aperevalov: I haven't seeing that but there's currently one patch for adding kuryr scenarios to rally
14:13:16 <dmellado> https://review.openstack.org/#/c/643562/
14:13:24 <dmellado> not there yet, but it'd be useful to check that once merged
14:14:01 <aperevalov> also I think we need to create blueprint regarding rebind linux kernel driver to uio/vfio.
14:16:03 <dmellado> aperevalov: feel free to create the blueprint and I'll take a look
14:16:16 <dmellado> regarding performance, since when? and how did you measure that?
14:18:08 <aperevalov> it's not our metrics, we didn't measure yet, our collogues complain, but rarely.
14:18:59 <dmellado> aperevalov: I'd be interested in checking this, would you mind getting the setup details so we could reproduce?
14:19:32 <aperevalov> they said, sometimes it takes about 30 seconds, since pod was applied till bind. So it could be bottleneck somewhere in our setup, I'm not sure it's in kuryr-kubernetes, but interesting did you faced that.
14:20:12 <dmellado> about to 30 seconds?
14:20:18 <dulek> I saw some binding performance issues when using OVN.
14:20:27 <dmellado> dulek: up to 30 seconds?
14:20:32 <aperevalov> dmellado, ok I'll try to gather more details.
14:20:40 <dulek> It takes a lot until bound port becomes ACTIVE in Neutron API.
14:20:41 <dmellado> aperevalov: sure, thanks!
14:20:45 <dulek> dmellado: 30-60 seconds.
14:21:21 <dmellado> dulek: that's just plain too much, what does OVN do on the meanwhile?
14:21:32 <dmellado> did you get to take a look at that by any chance?
14:22:38 <dulek> dmellado: I fixed one issue in OVN, but seems like it wasn't related.
14:22:56 <dulek> dmellado: So no, I don't know why it takes so long.
14:22:58 <dmellado> aperevalov: are you using ovn as a backend on your setup?
14:23:06 <dulek> But it's easily reproducible, just run DevStack with OVN.
14:23:12 <dmellado> dulek: all right, I'll try to set up an environment and engage the ovn folks
14:23:22 <dmellado> seems like a critical issue from my side
14:23:25 <aperevalov> no, we don't
14:24:30 <dmellado> in any case aperevalov, please check the setup details and we'll check ;)
14:25:02 <aperevalov> dmellado, ok
14:29:33 <dmellado> all right folks, then thanks for attending, we'll be at #openstack-kuryr
14:29:36 <dmellado> #endmeeting