| *** celebdor has quit IRC | 00:14 | |
| openstackgerrit | dongjie zhang proposed openstack/kuryr-libnetwork master: Modify "_get_fixed_ips_by_interface_cidr" https://review.openstack.org/552838 | 00:52 |
|---|---|---|
| *** salv-orlando has joined #openstack-kuryr | 00:55 | |
| *** salv-orlando has quit IRC | 01:00 | |
| *** salv-orlando has joined #openstack-kuryr | 01:56 | |
| *** salv-orlando has quit IRC | 02:01 | |
| *** dangtrinhnt has joined #openstack-kuryr | 02:52 | |
| *** hongbin has joined #openstack-kuryr | 02:57 | |
| *** salv-orlando has joined #openstack-kuryr | 02:57 | |
| *** salv-orlando has quit IRC | 03:01 | |
| *** hongbin has quit IRC | 03:57 | |
| *** salv-orlando has joined #openstack-kuryr | 03:58 | |
| *** salv-orlando has quit IRC | 04:02 | |
| *** gcheresh has joined #openstack-kuryr | 04:03 | |
| *** janki has joined #openstack-kuryr | 04:13 | |
| *** gcheresh has quit IRC | 04:38 | |
| *** janonymous has joined #openstack-kuryr | 04:50 | |
| *** salv-orlando has joined #openstack-kuryr | 04:59 | |
| *** salv-orlando has quit IRC | 05:03 | |
| *** gcheresh has joined #openstack-kuryr | 05:58 | |
| *** salv-orlando has joined #openstack-kuryr | 05:59 | |
| *** salv-orlando has quit IRC | 06:04 | |
| *** kiennt26 has joined #openstack-kuryr | 06:10 | |
| *** salv-orlando has joined #openstack-kuryr | 06:10 | |
| *** gcheresh has quit IRC | 06:17 | |
| *** maysamacedos has quit IRC | 06:33 | |
| *** yboaron has joined #openstack-kuryr | 07:07 | |
| *** maysamacedos has joined #openstack-kuryr | 07:36 | |
| *** celebdor has joined #openstack-kuryr | 08:14 | |
| *** maysamacedos has quit IRC | 08:28 | |
| *** maysamacedos has joined #openstack-kuryr | 08:28 | |
| *** jchhatbar has joined #openstack-kuryr | 08:32 | |
| *** janki has quit IRC | 08:32 | |
| ltomasbo | celebdor, I see you +W https://review.openstack.org/#/c/552856/ | 08:37 |
| ltomasbo | celebdor, what will happen when using tunneling? there will be a mismatch on the MTU size | 08:38 |
| celebdor | ltomasbo: not sure I understand | 08:38 |
| *** gcheresh has joined #openstack-kuryr | 08:38 | |
| ltomasbo | in case of having a default, to be on the safe side, I will set it to 1450 (so that vxlan and geneve will work properly) | 08:39 |
| celebdor | oh, that... | 08:39 |
| celebdor | Well, this is replicating the linux 1500 default | 08:39 |
| celebdor | but I would approve a follow-up patch to move the default to 1450 | 08:40 |
| celebdor | for the tunneling reason | 08:40 |
| ltomasbo | but the default in neutron is 1450 (for vxlan) | 08:40 |
| celebdor | ltomasbo: let me check a moment | 08:40 |
| celebdor | ltomasbo: I actually see 1442 with ovn (geneve) | 08:43 |
| celebdor | in tonicloud | 08:43 |
| ltomasbo | yes, geneve is a bit larger | 08:43 |
| ltomasbo | so, that is why yboaron and I were concern about setting mtu to 1500 | 08:43 |
| ltomasbo | it is tunneling dependent | 08:43 |
| celebdor | I understand | 08:43 |
| celebdor | ltomasbo: do you know where's the code that decides 1450/1442 in neutron | 08:44 |
| celebdor | ? | 08:44 |
| ltomasbo | and you can actually have different tunneling supported at the same time | 08:44 |
| ltomasbo | so some networks will have one, some other will have the other | 08:44 |
| *** yboaron has quit IRC | 08:44 | |
| ltomasbo | I think yboaron attached to his review a link to a document about that | 08:44 |
| ltomasbo | ihar was working on an automatic discovery of the mtu along the path | 08:45 |
| ltomasbo | problem is with old installation (when kuryr didn't exists) | 08:45 |
| ltomasbo | otherwise, the right mtu is at the network object | 08:45 |
| ltomasbo | celebdor, did you only enabled geneve? you can try to create a network with vxlan and see that the mtu is properly obtained | 08:46 |
| ltomasbo | (otherwise there is a bug) | 08:46 |
| celebdor | ltomasbo: I didn't know that ovn can work with vxlan | 08:57 |
| celebdor | can it? | 08:57 |
| ltomasbo | I meant that you can have geneve and vxlan tunneling enabled in neutron at the same time | 09:02 |
| ltomasbo | not sure if that is the great idea to make it working with ovn | 09:02 |
| ltomasbo | I think ovn has some support for vxlan, but not sure | 09:03 |
| ltomasbo | but I guess you can use geneve for ml2/ovs | 09:03 |
| *** gcheresh has quit IRC | 09:04 | |
| *** gcheresh has joined #openstack-kuryr | 09:07 | |
| celebdor | I have no ml2/ovs atm | 09:08 |
| *** jistr is now known as jistr|mtg | 09:24 | |
| *** kiennt26 has quit IRC | 10:24 | |
| *** gcheresh_ has joined #openstack-kuryr | 10:41 | |
| *** gcheresh has quit IRC | 10:43 | |
| celebdor | dulek: ping | 10:44 |
| dulek | celebdor: Pong. | 10:44 |
| celebdor | dulek: when you said that for cni side choice you only had bridge driver, did you mean ovs hybrid firewall bridge driver or native ovs? | 10:44 |
| dulek | celebdor: Should work for both, I think. I've only modified BaseBridgeDriver. | 10:45 |
| celebdor | ok | 10:45 |
| celebdor | dulek: 'reconnect' is what you called moving the iface from the dummy ns to the infra container one? | 10:46 |
| dulek | celebdor: Yup. | 10:46 |
| celebdor | :-) | 10:46 |
| celebdor | that's a funny name | 10:47 |
| dulek | celebdor: I should probably add a ton of docstrings. | 10:47 |
| celebdor | for a moment I thought it was checking if the dataplane went down somehow | 10:47 |
| celebdor | xD | 10:47 |
| dulek | celebdor: Naming is hard, you know. ;) | 10:47 |
| celebdor | indeed | 10:47 |
| dulek | celebdor: I can't even come up with something better. "move" doesn't sound good as well. | 10:47 |
| celebdor | dulek: where are we with the privsep stuff? | 10:48 |
| celebdor | does it still trip the queens release? | 10:48 |
| dulek | celebdor: Hm, why would it? | 10:48 |
| celebdor | well, we had that import issue, did we fix that? | 10:48 |
| celebdor | I forgot | 10:48 |
| dulek | celebdor: That was with os-vif>1.7.0. That was tracked down to usage of pyinstaller. | 10:49 |
| celebdor | ah, thanks | 10:49 |
| dulek | celebdor: So we're more-or-less fine at the moment - I've built upstream containers with older os-vif. | 10:49 |
| dulek | celebdor: And to fix that properly I want to move them to `docker run` CNI execution. | 10:50 |
| celebdor | yes, that part I remember | 10:50 |
| celebdor | I think for upstream we can probably go use runc directly | 10:50 |
| dulek | celebdor: I've planned this for today, but currently I'm trying to tackle https://review.openstack.org/#/c/544548 - to unblock Neutron folks. | 10:50 |
| celebdor | dulek: yes, yes. Unblock neutron | 10:51 |
| celebdor | for now I'm reviewing your patch | 10:51 |
| dulek | celebdor: I appreciate that, I definitely need some help there. | 10:51 |
| dulek | This daemon-side VIF choice is easy in the big picture, but mind-numbing in details. At least to me. :P | 10:52 |
| celebdor | ;-) | 10:53 |
| openstackgerrit | Merged openstack/kuryr-kubernetes master: Deprecate running kuryr-k8s without kuryr-daemon https://review.openstack.org/550170 | 10:58 |
| *** gcheresh_ has quit IRC | 11:02 | |
| *** jistr|mtg is now known as jistr | 11:07 | |
| celebdor | dulek: any reason to put the different plugins in the same module https://review.openstack.org/#/c/527243/11/kuryr_kubernetes/cni/plugins/k8s_cni_registry.py ? | 11:18 |
| dulek | celebdor: Not really, but both share this base registry plugin, so it was easier to follow when debugging. Once finished those should be split. | 11:19 |
| celebdor | ok | 11:19 |
| celebdor | dulek: why are you annotating the kuryrport with the podname? | 11:23 |
| celebdor | Isn't it better to just annotate the pod with the kuryrport? | 11:24 |
| *** yamamoto has quit IRC | 11:24 | |
| dulek | celebdor: Yup, I haven't yet implemented what we've came up with during the PTG. | 11:24 |
| celebdor | ok | 11:24 |
| celebdor | dulek: Do we need to watch for pod on present? | 11:26 |
| celebdor | rather, do we need to watch for pod events? | 11:27 |
| dulek | celebdor: Uhm, interesting question… | 11:27 |
| celebdor | I would have imagined that once we are on cni side | 11:28 |
| celebdor | we only react to KuryrPort creation to activate | 11:28 |
| dulek | celebdor: Yup, yup, I get it, CNI ADD/DEL. | 11:28 |
| celebdor | and then on CNI_ADD | 11:28 |
| celebdor | exactly | 11:28 |
| celebdor | basically we start behaving a bit like a thin cni plugin | 11:28 |
| dulek | celebdor: I don't immediately see any blockers, let me glance into the code for a moment. | 11:29 |
| celebdor | ok | 11:30 |
| dulek | celebdor: InstanceInfo passed to os.vif can be mocked with CNI params pretty easily, so it's not an issue. | 11:32 |
| dulek | celebdor: But getting the pool_key to know where to look for VIF is an issue. | 11:33 |
| dulek | celebdor: I mean not now, when we only have SGs and project as config values. | 11:33 |
| dulek | celebdor: But once those will be dependent on pod - this might be a problem. | 11:33 |
| dulek | celebdor: Of course we can just fetch pod info on demand from K8s API. | 11:34 |
| celebdor | mmm | 11:34 |
| celebdor | but those would be for namespace, wouldn't they | 11:34 |
| celebdor | and we know hte namespace on CNI_ADD | 11:34 |
| celebdor | or do you mean for the network policy? | 11:34 |
| dulek | celebdor: SGs and network policies would be an issue. | 11:35 |
| dulek | celebdor: You're right on namespace thing, though. :) | 11:35 |
| celebdor | The more I think about it, the more it seems to me like that for network policy we need the controller to assign once the kuryrport has been assigned | 11:36 |
| celebdor | but that comes with some assumptions | 11:36 |
| dulek | celebdor: We can just fetch it on demand when we're running with network policies. It'll be a tremendous performance hit anyway. | 11:36 |
| celebdor | yeah | 11:36 |
| openstackgerrit | Michał Dulko proposed openstack/kuryr-libnetwork master: Check for `standard-attr-tag` Neutron extension https://review.openstack.org/553756 | 11:37 |
| openstackgerrit | Michał Dulko proposed openstack/kuryr-libnetwork master: Check for `standard-attr-tag` Neutron extension https://review.openstack.org/553756 | 11:53 |
| openstackgerrit | Michał Dulko proposed openstack/kuryr-libnetwork master: DNM: Testing a standard-attr-tag fix https://review.openstack.org/553763 | 11:53 |
| *** yamamoto has joined #openstack-kuryr | 11:58 | |
| *** gcheresh_ has joined #openstack-kuryr | 11:58 | |
| *** yamamoto has quit IRC | 12:31 | |
| *** yamamoto has joined #openstack-kuryr | 12:36 | |
| *** moycavirp has joined #openstack-kuryr | 12:39 | |
| *** maysamacedos has quit IRC | 12:44 | |
| *** jchhatbar has quit IRC | 12:47 | |
| *** jchhatba_ has joined #openstack-kuryr | 12:47 | |
| *** jlabarre has quit IRC | 12:58 | |
| *** jlabarre has joined #openstack-kuryr | 13:00 | |
| *** salv-orlando has quit IRC | 13:01 | |
| *** salv-orlando has joined #openstack-kuryr | 13:17 | |
| *** janonymous has quit IRC | 13:18 | |
| *** caowei has quit IRC | 13:21 | |
| dulek | celebdor: This `runc` thing doesn't look too pretty: https://gist.github.com/mitake/0793dd07bef3d8170fe979774db829c3 | 13:26 |
| dulek | celebdor: Is that what you've meant? | 13:26 |
| celebdor | not this flow | 13:28 |
| celebdor | but it's the same tool | 13:28 |
| dulek | celebdor: What's the correct flow then? | 13:29 |
| celebdor | runc exec | 13:30 |
| celebdor | just like you'd do docker exec | 13:30 |
| dulek | celebdor: Ah, okay. So you assume container is up already. | 13:30 |
| celebdor | yes | 13:31 |
| celebdor | can't we | 13:31 |
| celebdor | ? | 13:31 |
| dulek | celebdor: We sure can, but that would be… `docker run`? Or you mean to start it with runc as well? | 13:31 |
| celebdor | dulek: the container is started by kubelect since its the daemonset | 13:33 |
| celebdor | docker uses runc to start it | 13:33 |
| celebdor | (for a recent enough version of docker) | 13:33 |
| dulek | celebdor: Oooh… | 13:34 |
| celebdor | and then we just do runc exec | 13:34 |
| dulek | celebdor: Okay, I'm not sure why I haven't thought of it. I'm glad I asked. :) | 13:34 |
| celebdor | :-) | 13:34 |
| *** jchhatba_ has quit IRC | 13:35 | |
| celebdor | dulek: http://paste.openstack.org/show/703010/ | 13:37 |
| dulek | celebdor: Do you know how id is calculated? | 13:40 |
| dulek | celebdor: I can inject into container's env anything that's in container spec. I wonder if there's anything that matches. | 13:41 |
| dulek | celebdor: BTW - what versions of Docker and runc do you have? My runc had issues with container file structure. | 13:45 |
| celebdor | what do you need spec? | 13:46 |
| celebdor | the container is still built with docker | 13:46 |
| celebdor | we would only use runc for doing the exec | 13:46 |
| dulek | celebdor: http://paste.openstack.org/show/703012/ | 13:47 |
| celebdor | I have runc 0.1.1 | 13:48 |
| celebdor | and docker 17.05.0-ce | 13:48 |
| *** hongbin has joined #openstack-kuryr | 14:00 | |
| *** vikasc has quit IRC | 14:03 | |
| *** vikasc has joined #openstack-kuryr | 14:08 | |
| *** hongbin has quit IRC | 14:20 | |
| *** hongbin has joined #openstack-kuryr | 14:20 | |
| *** caowei has joined #openstack-kuryr | 14:28 | |
| *** caowei has quit IRC | 14:29 | |
| *** janki has joined #openstack-kuryr | 15:25 | |
| *** yamamoto has quit IRC | 15:26 | |
| *** yamamoto has joined #openstack-kuryr | 15:36 | |
| *** yamamoto has quit IRC | 15:41 | |
| *** yamamoto has joined #openstack-kuryr | 15:51 | |
| *** yamamoto has quit IRC | 15:55 | |
| *** jistr_ has joined #openstack-kuryr | 16:01 | |
| *** jistr has quit IRC | 16:03 | |
| *** celebdor has quit IRC | 16:03 | |
| *** jistr_ is now known as jistr | 16:04 | |
| *** yamamoto has joined #openstack-kuryr | 16:06 | |
| *** yamamoto has quit IRC | 16:11 | |
| *** yamamoto has joined #openstack-kuryr | 16:21 | |
| *** yamamoto has quit IRC | 16:25 | |
| *** yamamoto has joined #openstack-kuryr | 16:36 | |
| *** yamamoto has quit IRC | 16:41 | |
| *** yamamoto has joined #openstack-kuryr | 16:51 | |
| *** yamamoto has quit IRC | 16:56 | |
| *** neiljerram has quit IRC | 17:06 | |
| *** yamamoto has joined #openstack-kuryr | 17:06 | |
| *** yamamoto has quit IRC | 17:07 | |
| *** yamamoto has joined #openstack-kuryr | 17:07 | |
| *** yamamoto has quit IRC | 17:07 | |
| *** gcheresh_ has quit IRC | 17:14 | |
| *** celebdor has joined #openstack-kuryr | 17:49 | |
| *** vikasc has quit IRC | 17:51 | |
| *** yamamoto has joined #openstack-kuryr | 18:08 | |
| *** yamamoto has quit IRC | 18:14 | |
| *** vikasc has joined #openstack-kuryr | 18:18 | |
| *** gcheresh_ has joined #openstack-kuryr | 18:45 | |
| *** gcheresh_ has quit IRC | 18:53 | |
| *** gcheresh has joined #openstack-kuryr | 19:00 | |
| *** janki has quit IRC | 19:05 | |
| *** celebdor has quit IRC | 19:06 | |
| *** yamamoto has joined #openstack-kuryr | 19:10 | |
| *** yamamoto has quit IRC | 19:20 | |
| *** celebdor has joined #openstack-kuryr | 19:30 | |
| *** gcheresh has quit IRC | 19:54 | |
| *** maysamacedos has joined #openstack-kuryr | 20:15 | |
| *** yamamoto has joined #openstack-kuryr | 20:17 | |
| *** yamamoto has quit IRC | 20:22 | |
| *** yamamoto has joined #openstack-kuryr | 21:18 | |
| *** gcheresh has joined #openstack-kuryr | 21:21 | |
| *** maysamacedos has quit IRC | 21:22 | |
| *** yamamoto has quit IRC | 21:23 | |
| *** gcheresh has quit IRC | 21:29 | |
| *** vikasc has quit IRC | 22:07 | |
| *** hongbin has quit IRC | 22:07 | |
| *** mestery has quit IRC | 22:07 | |
| *** olivierbourdon38 has quit IRC | 22:07 | |
| *** vikasc has joined #openstack-kuryr | 22:07 | |
| *** hongbin has joined #openstack-kuryr | 22:07 | |
| *** mestery has joined #openstack-kuryr | 22:07 | |
| *** olivierbourdon38 has joined #openstack-kuryr | 22:07 | |
| *** salv-orlando has quit IRC | 22:15 | |
| *** salv-orlando has joined #openstack-kuryr | 22:16 | |
| *** yamamoto has joined #openstack-kuryr | 22:20 | |
| *** yamamoto has quit IRC | 22:26 | |
| *** hongbin has quit IRC | 22:48 | |
| *** maysamacedos has joined #openstack-kuryr | 23:20 | |
| *** yamamoto has joined #openstack-kuryr | 23:22 | |
| *** yamamoto has quit IRC | 23:27 | |
| *** maysamacedos has quit IRC | 23:45 | |
| *** openstackgerrit has quit IRC | 23:48 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!