14:00:39 <vikasc> #startmeeting kuryr 14:00:40 <openstack> Meeting started Mon Jan 23 14:00:39 2017 UTC and is due to finish in 60 minutes. The chair is vikasc. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:42 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:44 <openstack> The meeting name has been set to 'kuryr' 14:00:54 <vikasc> Hello and welcome everybody to another Kuryr weekly IRC meeting 14:00:58 <janonymous> o/ 14:01:03 <irenab> hi 14:01:04 <garyloug> o/ 14:01:05 <ivc_> o/ 14:01:16 <vikasc> today apuimedo|flu is down with flu 14:01:19 <mchiappero> o/ 14:01:25 <alraddarla> o/ 14:01:36 <limao> o/ 14:01:57 <irenab> shall we start? 14:02:14 <vikasc> #info janki irenab ivc_ garyloug mchiappero alraddarla limao vikasc joined the meeting 14:02:23 <vikasc> irenab, sure 14:02:26 <janonymous> +1 14:02:29 <janonymous> :P 14:02:33 <ltomasbo> o/ 14:02:34 <vikasc> #topic kuryr-lib 14:02:40 <reedip_> hi 14:02:43 <vikasc> :) 14:03:01 <vikasc> #link https://review.openstack.org/#/c/422641/ 14:03:26 <vikasc> ltomasbo added a patch for adding randomness to segmentation driver 14:04:04 <janonymous> i need help with: https://review.openstack.org/#/c/410609/ i guess :O 14:04:06 <vikasc> looks like it is almost ready to merge, limao reviewed it extensivly 14:04:22 <irenab> vikasc: we probably need to decide when to cut new kuryr lib version 14:04:41 <janki> hi 14:04:52 <vikasc> janonymous, help in like? 14:04:53 <irenab> janonymous: I will look into the patch after the meeting 14:05:21 <janonymous> vikasc: some problem on docker side with tls 14:05:24 <vikasc> #info janonymous has a patch for tls support 14:05:29 <janonymous> irenab:thanks 14:05:35 <vikasc> #link https://review.openstack.org/#/c/410609/ 14:05:43 <vikasc> hi janki 14:06:02 <vikasc> thanks irenab 14:06:09 <janonymous> vikasc:thanks 14:06:36 <vikasc> irenab, yes, on version 14:06:39 <irenab> janonymous: please post the issues/concerns you have as comment on the patch 14:07:09 <janonymous> limao: mentioned the problem in comment, thanks to him :) 14:07:37 <vikasc> irenab, lets target port merging ltomasbo patch for randomness, WDYT? 14:07:38 <irenab> vikasc: I am not sure if there is some policy in place, we should decide when the changes are meaningful enough to cut the version 14:08:15 <irenab> vikasc: sure 14:08:41 <vikasc> #action to discuss on policy for cutting kuryr-lib release version 14:08:52 <vikasc> ok lets move on 14:09:10 <vikasc> do we have anything else on kuryr lib? 14:09:27 <vikasc> #topic kuryr-libnetwork 14:09:36 <irenab> vikasc: the binding patch with mtu? 14:10:21 <vikasc> irenab, that i was thinking to discuss in kuryr-k8s section, since we decided that we will add any new changes in specific rpos 14:10:24 <vikasc> s/repos 14:10:33 <irenab> vikasc: ok 14:10:40 <vikasc> and then get them bublled up to kuryr-lib 14:10:51 <vikasc> irenab, thanks 14:11:11 <ivc_> dont we already have mtu in k8s? 14:11:16 <vikasc> we merged a patch for code cleanup 14:11:19 <vikasc> #link https://review.openstack.org/#/c/415363/ 14:12:08 <vikasc> ivc_, we have, it is a specif case 14:12:31 <vikasc> ivc_, i will explain more in a while 14:12:48 <irenab> vikasc: shall we move to k8s topic? 14:12:54 <vikasc> #link https://review.openstack.org/#/c/424067/ 14:13:13 <vikasc> ltomasbo added a patch for a fix in subport deletion 14:13:22 <vikasc> irenab, sure.. just after this 14:13:49 <vikasc> ltomasbo, would you like to share discussion/findings you had on this 14:13:59 <ltomasbo> yep, it seems using device_owner is not a good approach 14:14:09 <ltomasbo> as it could collide with some other usage at neutron side 14:14:18 <ltomasbo> so, it is proposed to use tagging instead 14:14:28 <ltomasbo> but until tags are supported at neutron for ports 14:14:53 <irenab> looks like neutron does not guarantee to preserve device_owner for trunk service case 14:15:12 <ltomasbo> we need to remove the filtering by device_owner, as it may leave ports not deleted for vlan case 14:15:21 <ivc_> we could also take another approach by querying trunks instead of sub-ports 14:15:49 <ltomasbo> yep, neutron tries to use device_owner for trunk subports, and that colides with our usage 14:15:52 <ivc_> tho it would make sub-ports handling 'special' and completely different from normal ports 14:16:32 <ivc_> but that 'list_port' in ltomasbo patch worries me in case we have some large deployment 14:16:39 <irenab> ivc_: seems that tags support for ports is in progress, better avoid special cases on kuryr side 14:17:02 <irenab> ivc_: large and mixed deployments 14:17:11 <ltomasbo> I agree with irenab, plus device_owner is just used in kuryr-libnetwork for removing the subports 14:17:39 <ivc_> irenab ltomasbo maybe it's time for us to consider some sort of our own storage for kuryr-related data 14:17:50 <vikasc> may be it makes more sense to wait till support is ready on neutron side 14:18:03 <ivc_> tags are sort of close to device_owner in terms of their reliability 14:18:04 <irenab> ivc_: so hoped to keep kuryr stateless ;-) 14:18:29 <ivc_> irenab well in kuryr-k8s we have k8s as backing store, so not so stateless 14:18:56 <ivc_> i mean annotations are our 'storage' 14:19:12 <vikasc> ok, lets move on to kuryr-k8s section then :) 14:19:21 <vikasc> #topic kuryr-kubernetes 14:19:55 <vikasc> #link https://review.openstack.org/#/c/376045/ 14:20:16 <vikasc> ivc_, seems you are segregating services patch in several smaller patches 14:20:24 <ivc_> aye 14:20:35 <irenab> ivc_: I need to give storage topic some thoughts, but we should go for it if its unavoidable 14:20:40 <hongbin> o/ 14:20:48 <vikasc> hello hongbin 14:21:19 <vikasc> irenab, we can add this to topics for virtual gathering 14:21:50 <ivc_> there are few patches cut from https://review.openstack.org/#/c/376045/ that are not directly related to services but are small improvements 14:21:54 <irenab> vikasc: +1 14:21:59 <vikasc> #info we merged nested-pods support in kuryr-k8s 14:22:15 <vikasc> #link https://review.openstack.org/#/c/410578/ 14:22:34 <irenab> sorry, need to leave the meeting now 14:22:44 <vikasc> ivc_, very nice, will go yhrough those 14:22:54 <vikasc> irenab, thanks for joining!! 14:23:00 <vikasc> s/through 14:23:27 <vikasc> #link https://review.openstack.org/#/c/421777/3 14:23:59 <vikasc> we merged an improvement patch in exception handling by apuimedo|flu 14:24:37 <vikasc> i think thats pretty much all from kuryr-k8s 14:24:53 <vikasc> ahh sorry, missed mtu discussion 14:24:58 <ivc_> yup 14:25:24 <vikasc> #link https://review.openstack.org/#/c/424104/ 14:26:04 <vikasc> there might be an issue if user manually decreases mtu inside vm on vm interface 14:26:57 <ivc_> maybe we need a warning for that 14:27:02 <vikasc> when kuryr-cni try to update the mtu on child device, pyroute2 will throw an error because mtu will be more than what is there on vm interface 14:27:37 <ivc_> actually we do need a warning. if the user messed up the vm config and reduced mtu, it could lead to performance degradation 14:27:53 <mchiappero> do you really need to force MTU? 14:28:00 <mchiappero> I'm just wondering 14:28:18 <ivc_> mchiappero force to what? 14:28:25 <vikasc> i too dint get 14:28:26 <ivc_> to neutron value or to the vm value? 14:28:52 <vikasc> we need to for vm value because of pyroute2 constraint 14:29:03 <mchiappero> maybe I misunderstood but where is this MTU value coming from? 14:29:18 <vikasc> one value is on vm interface 14:29:26 <ivc_> vikasc i think that error is actually good to have 14:29:42 <ivc_> i'd prefer the user to see the error right away 14:29:42 <vikasc> and other is coming from neutron which kuryr-cni is updating in vlan device 14:30:17 <ivc_> otherwise we'll silently change mtu to a less-then-optimal value that would leave the user wondering about less-then-optimal performance 14:30:23 <mchiappero> ok, what I meant was: can't you just ignore the one from Neutron? 14:30:23 <ivc_> fail fast (tm) 14:30:33 <vikasc> may be we will take this up further in open discussion, please feel free to add review comments everybody 14:31:01 <ivc_> vikasc sure 14:31:01 <vikasc> thats what has been done in the patch 14:31:10 <mchiappero> oh, ok 14:31:29 <vikasc> hongbin would you like to share an update on fuxi 14:31:35 <vikasc> #topic fuxi 14:31:36 * ivc_ is against ignoring neutron mtu :) 14:31:36 <hongbin> vikasc: sure 14:31:43 <vikasc> #chair hongbin 14:31:43 <openstack> Current chairs: hongbin vikasc 14:31:47 <hongbin> i don't have too much last week 14:31:59 <hongbin> some contributors submitted some patches last week 14:32:07 <hongbin> #link https://review.openstack.org/#/c/423570/ 14:32:12 <hongbin> this one for fixing the logging 14:32:19 <hongbin> #link https://review.openstack.org/#/c/415378/ 14:32:29 <hongbin> this one for improving the config group 14:32:52 <hongbin> that is it from last week, i will start working on the kolla integration next week 14:33:02 <hongbin> vikasc: that is all from my side 14:33:06 <vikasc> very cool! 14:33:14 <vikasc> thanks hongbin !! 14:33:25 <vikasc> #topic open-discussion 14:34:31 <vikasc> anybody whats to share any idea , please feel free to come in 14:35:10 <vikasc> #info virtual gathering etherpad, https://etherpad.openstack.org/p/kuryr_virtual_gathering_2017h1 14:35:30 <vikasc> guys please take a look and update the details 14:36:27 <vikasc> anything else to add on.. 14:37:10 <vikasc> cool.. 14:37:30 <vikasc> seems we need to wrap up 14:37:38 <vikasc> thanks evrybody !!! 14:37:43 <vikasc> #endmeeting