| janonymous | limao: hey | 00:05 |
|---|---|---|
| *** limao has joined #openstack-kuryr | 00:47 | |
| *** saneax is now known as saneax-_-|AFK | 00:53 | |
| *** yedongcan has joined #openstack-kuryr | 01:09 | |
| *** yedongcan has quit IRC | 02:21 | |
| *** vikasc has joined #openstack-kuryr | 02:54 | |
| *** yedongcan has joined #openstack-kuryr | 03:11 | |
| *** limao has quit IRC | 03:40 | |
| *** limao has joined #openstack-kuryr | 03:42 | |
| *** limao has quit IRC | 03:47 | |
| *** yedongcan1 has joined #openstack-kuryr | 03:51 | |
| *** yedongcan has quit IRC | 03:52 | |
| *** reedip has quit IRC | 04:35 | |
| *** limao has joined #openstack-kuryr | 04:42 | |
| openstackgerrit | Ilya Chukhnakov proposed openstack/kuryr-kubernetes: Improve pipeline/Async logging https://review.openstack.org/423903 | 04:43 |
| *** v1k0d3n has quit IRC | 04:44 | |
| *** v1k0d3n has joined #openstack-kuryr | 04:44 | |
| *** limao has quit IRC | 04:47 | |
| *** reedip has joined #openstack-kuryr | 04:47 | |
| *** limao has joined #openstack-kuryr | 05:19 | |
| *** limao has quit IRC | 05:23 | |
| *** limao has joined #openstack-kuryr | 05:25 | |
| *** hongbin has quit IRC | 05:28 | |
| openstackgerrit | Ilya Chukhnakov proposed openstack/kuryr-kubernetes: OVO model for K8s Services support https://review.openstack.org/423908 | 05:33 |
| *** v1k0d3n has quit IRC | 05:56 | |
| *** reedip has quit IRC | 05:58 | |
| *** v1k0d3n has joined #openstack-kuryr | 06:02 | |
| *** reedip has joined #openstack-kuryr | 06:10 | |
| *** saneax-_-|AFK is now known as saneax | 06:19 | |
| *** janki has joined #openstack-kuryr | 06:26 | |
| *** gsagie has joined #openstack-kuryr | 06:58 | |
| *** yamamoto has quit IRC | 08:10 | |
| *** pcaruana has joined #openstack-kuryr | 08:17 | |
| openstackgerrit | Jaivish Kothari(janonymous) proposed openstack/kuryr-libnetwork: Tls support configurations https://review.openstack.org/410609 | 08:23 |
| *** yedongcan1 has quit IRC | 08:28 | |
| *** yedongcan has joined #openstack-kuryr | 08:29 | |
| *** yamamoto has joined #openstack-kuryr | 08:46 | |
| openstackgerrit | Anh Tran proposed openstack/kuryr-libnetwork: Typo fix: happend => happened https://review.openstack.org/423995 | 08:54 |
| *** limao has quit IRC | 09:12 | |
| *** limao has joined #openstack-kuryr | 09:16 | |
| *** yamamoto has quit IRC | 09:21 | |
| *** limao has quit IRC | 09:21 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr: Add randomness to the returned vlan_ids https://review.openstack.org/422641 | 09:25 |
| *** yamamoto has joined #openstack-kuryr | 09:40 | |
| *** devvesa has joined #openstack-kuryr | 09:58 | |
| *** yedongcan has left #openstack-kuryr | 10:07 | |
| *** apuimedo is now known as apuimedo|flu | 10:13 | |
| *** irenab has left #openstack-kuryr | 10:14 | |
| *** irenab has joined #openstack-kuryr | 10:15 | |
| ltomasbo | apuimedo|flu, irenab, vikasc, all: I've been discussing with armax the device_owner usage | 10:24 |
| ltomasbo | https://review.openstack.org/#/c/419028/ | 10:24 |
| ltomasbo | he suggest if we can use tagging instead of the device_owner | 10:24 |
| irenab | ltomasbo: is tagging on other than network resource got merged? | 10:24 |
| ltomasbo | as it seems there is no consensus about how this field should be used, and could lead to problems (as the one we are hitting)( | 10:25 |
| ltomasbo | not sure, don't think so: https://review.openstack.org/#/c/419028/ | 10:25 |
| ltomasbo | I checked and we only use device_owner in once in kuryr-libnetwork | 10:26 |
| irenab | just saw it, https://review.openstack.org/#/c/413662 | 10:26 |
| irenab | not tags to use yet, but can be a way to go | 10:26 |
| ltomasbo | just to filter out some devices before removing the port when deleting the container | 10:26 |
| irenab | ltomasbo: the meaning of the device_owner field in neutron port is really not clear. Tags will be better since it is meant to be for the external managment usage | 10:28 |
| ltomasbo | yep, I think it will be a safer choice | 10:29 |
| ltomasbo | and I just checked, we only use device_owner once in kuryr-libnetwork | 10:29 |
| ltomasbo | and not at all in kuryr-kubernetes | 10:29 |
| ltomasbo | so, maybe, until tags are merged in neutron, I can propose a fix to the kuryr-libnetwork, and we can also work in parallel to set tags instead of device_owner | 10:30 |
| irenab | ltomasbo: this is good. I think we can set it, but definetly not count on it for some operations later. Seems there is no contract on keeping the provided value on create | 10:30 |
| *** neiljerram has joined #openstack-kuryr | 10:31 | |
| ltomasbo | you mean for tagging? | 10:31 |
| irenab | ltomasbo: device_owner field. Tags are meant to be set by external to neutron entity, so the contract is that neutron won’t change it, actually it is not supposed to pass it to the back-ends even | 10:32 |
| ltomasbo | so, even more in favor of using tags instead of device_owner... | 10:33 |
| irenab | yes | 10:33 |
| ltomasbo | great! thanks! | 10:33 |
| irenab | we just need to rely on the patch https://review.openstack.org/#/c/413662 | 10:33 |
| ltomasbo | yep, I will reply to armax, and work on that | 10:36 |
| *** jchhatbar has joined #openstack-kuryr | 10:43 | |
| *** janki has quit IRC | 10:45 | |
| *** portdirect_away is now known as portdirect | 10:59 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-libnetwork: Ensure subports are deleted after container deletion https://review.openstack.org/424067 | 11:27 |
| *** yamamoto has quit IRC | 11:34 | |
| vikasc | irenab, ltomasbo , i want to know you opinion on handling one scenario | 11:48 |
| irenab | vikasc: go ahead | 11:48 |
| vikasc | in nested-pods case, if mtu of vm interface is less than what binding code has fetched from neutron port and try to set it on container interface, it fails | 11:49 |
| vikasc | i was just playing around. my vm has 1400 mtu and neutron port has 1450.. so in binding when it tries to set 1450 on container port and try to iface.up() it fails | 11:50 |
| vikasc | irenab, ltomasbo ^ | 11:51 |
| ltomasbo | umm | 11:51 |
| vikasc | i verified that if neutron_port mtu <= vm interface mtu, no problem | 11:51 |
| ltomasbo | and what is the MTU of the server nic? | 11:53 |
| irenab | vikasc: the trunk port mtu < subport mtu? | 11:53 |
| ltomasbo | I guess the subport is using the same vNIC as the VM, so the MTU should be the same | 11:54 |
| vikasc | ltomasbo, server means vm? its 1400 | 11:54 |
| vikasc | irenab, yes this is the error case | 11:54 |
| vikasc | ltomasbo, it should be same ideally, but sometimes vms are not able to work with neutron advertised mtu and user may need to lower it | 11:55 |
| irenab | vikasc: sounds like the subport should be less than a trunk | 11:55 |
| vikasc | irenab, will it be reasonable to add a check in binding to ensure this | 11:56 |
| irenab | vikasc: on kuryr side? | 11:56 |
| vikasc | irenab, yes | 11:56 |
| ltomasbo | shouldn't this be ensure at trunkport functionality? | 11:56 |
| irenab | it maybe should go by the lowest | 11:56 |
| irenab | ltomasbo: good question | 11:57 |
| vikasc | irenab, yes.. this check.. | 11:57 |
| ltomasbo | maybe it is, unless you specify a different one... | 11:57 |
| ltomasbo | going to check | 11:57 |
| irenab | vikasc: it maybe not even for a check sake, for getting the MTU set properly | 11:57 |
| vikasc | ltomasbo, this problem will occur only if user has manuaaly lowered the vm interface mtu | 11:58 |
| irenab | so this is something that cannot be checked on the api request handling | 11:59 |
| vikasc | irenab, i do not think it can be checked. i give you my example | 11:59 |
| irenab | vikasc: the option for figuring out the MTU on kuryr side looks reasonable to me | 11:59 |
| ltomasbo | https://github.com/openstack/neutron/blob/master/neutron/services/trunk/rules.py#L172:L191 | 12:00 |
| vikasc | irenab, ltomasbo , i prepared setup and neutron setted 1450 as mtu on vm | 12:00 |
| irenab | vikasc: so it retrieves the subport network mtu, correct? | 12:01 |
| ltomasbo | are you using vxlan? | 12:01 |
| ltomasbo | it is good to have 1450 to allow vxlan headers | 12:01 |
| vikasc | irenab, yes | 12:01 |
| vikasc | ltomasbo, yes | 12:01 |
| irenab | vikasc: the method below, there is a validation for the case we discuss | 12:02 |
| irenab | https://github.com/openstack/neutron/blob/master/neutron/services/trunk/rules.py#L196 | 12:02 |
| vikasc | ltomasbo, but for my vm, 1450 was not working, i had to lower it | 12:02 |
| vikasc | ltomasbo, irenab, ssl connection to a docker repo was hanging | 12:03 |
| ltomasbo | ummm, weird | 12:03 |
| vikasc | ltomasbo, irenab i struggled a lot to find out that if i lower the mtu of vm , it started working | 12:03 |
| irenab | vikasc: interesting | 12:04 |
| ltomasbo | but for the parent port? | 12:04 |
| vikasc | ltomasbo, from inside the vm | 12:04 |
| ltomasbo | that seems like a bug somewhere | 12:04 |
| vikasc | for the vm port only i had to lower the mtu to get connection working with the docker registry | 12:04 |
| ltomasbo | I've never had problems with the MTU for the trunkports | 12:04 |
| ltomasbo | only with the docker registry? | 12:05 |
| ltomasbo | the other connectivity was working well? | 12:05 |
| vikasc | ltomasbo, irenab , not sure, but my understanding is that it depends on if there is a network switch in the path that has lower mtu | 12:05 |
| vikasc | ltomasbo, yes | 12:05 |
| vikasc | ltomasbo, so problem may depend on destination from same source(my vm) | 12:06 |
| vikasc | ltomasbo, i can show you over bj :) | 12:06 |
| vikasc | ltomasbo, now only if you are interested | 12:06 |
| vikasc | irenab, i will add a check then to handle such problem | 12:07 |
| vikasc | irenab, i was struggling since last 2-3 hours to figure out why binding code not working :) | 12:07 |
| irenab | vikasc: I agree with your direction to make binding code more robust | 12:09 |
| vikasc | irenab, error that pyroute2 was throwing was not saying anything about mtu .. it was saying "index out of range" :) | 12:09 |
| irenab | vikasc: sounds like bug for pyroute2 :-) | 12:09 |
| *** jchhatbar_ has joined #openstack-kuryr | 12:10 | |
| *** jchhatbar_ is now known as janki | 12:10 | |
| vikasc | irenab, will raise it on pyroute2 and add in commit message for reference | 12:10 |
| *** jchhatbar has quit IRC | 12:13 | |
| ltomasbo | vikasc, I cannot connect by bj right now, but may ping you later or tomorrow | 12:14 |
| ltomasbo | btw, I manage to create nested containers with your patch after moving to a slightly bigger VM | 12:15 |
| vikasc | ltomasbo, sure.. anytime :) | 12:15 |
| ltomasbo | at the end it was a cpu + timeout problem, not just memory | 12:15 |
| vikasc | oh | 12:15 |
| vikasc | and it was kiing handler? | 12:16 |
| vikasc | s/killing | 12:16 |
| vikasc | or handler was timing out, starved on cpu? | 12:16 |
| ltomasbo | time oyut | 12:18 |
| ltomasbo | time out | 12:18 |
| ltomasbo | it took more than 20 minutes to boot up the container the first time | 12:19 |
| irenab | ltomasbo: vikasc : we probably need to add recommended flavor and image in the readme section before we have fullstack tests | 12:19 |
| ltomasbo | so I moved to another deployment with a bit more memory and then it worked | 12:19 |
| irenab | ltomasbo: I had similar issues | 12:19 |
| ltomasbo | irenab, yes, definitely agreed | 12:19 |
| ltomasbo | the 'overcloud' should have at least 4 GB and 2 vcpus | 12:20 |
| vikasc | true | 12:20 |
| vikasc | ltomasbo, irenab shall i add this detail to readme? | 12:21 |
| irenab | I wonder if we could have small image just for kuryr overcloud and not entire fedora/ubuntu | 12:21 |
| irenab | vikasc: yes, definitely | 12:21 |
| vikasc | irenab, but then k8s is also needed | 12:23 |
| vikasc | irenab, for a all-in-one setup | 12:23 |
| ltomasbo | maybe it could be usefull to have kuryr-controller in one VM | 12:23 |
| ltomasbo | and kuryr-cni in the worker node | 12:23 |
| ltomasbo | perhaps we can then simplify the VM requirement (although 2 will be needed, but more close to real deployments) | 12:24 |
| vikasc | ltomasbo, but our readme instructions are for single node only | 12:24 |
| ltomasbo | vikasc, yep, you are right, let's keep it simple | 12:24 |
| vikasc | ltomasbo, may be we can add 4GB and 2 vcpu for all-in-one kind | 12:25 |
| vikasc | and then user can accordingly estimate if he goes for multinode | 12:25 |
| ltomasbo | that is what I needed for the overcloud VM | 12:25 |
| vikasc | anyways, starvation will be less on multinode due to load distribution | 12:25 |
| ltomasbo | yes vikasc, keep it simple, maybe just mention that is for single-node, and that it could be splitted in different VMs if needed | 12:26 |
| vikasc | ltomasbo, make sense!! | 12:26 |
| ltomasbo | irenab, vikasc: btw, can you look at this two: https://review.openstack.org/#/c/424067/ https://review.openstack.org/#/c/422641/ | 12:27 |
| ltomasbo | s/this/these | 12:27 |
| vikasc | ltomasbo, asap | 12:28 |
| ltomasbo | thanks! | 12:28 |
| *** yamamoto has joined #openstack-kuryr | 12:34 | |
| openstackgerrit | vikas choudhary proposed openstack/kuryr-kubernetes: Update ReadMe for nested-pods setup resource requirements https://review.openstack.org/424093 | 12:48 |
| *** saneax is now known as saneax-_-|AFK | 12:53 | |
| *** dougbtv has joined #openstack-kuryr | 13:07 | |
| *** yamamoto has quit IRC | 13:14 | |
| openstackgerrit | vikas choudhary proposed openstack/kuryr-kubernetes: Handle mtu on kuryr side https://review.openstack.org/424104 | 13:15 |
| leifmadsen | o/ | 13:15 |
| *** v1k0d3n has quit IRC | 13:28 | |
| *** v1k0d3n has joined #openstack-kuryr | 13:29 | |
| *** saneax-_-|AFK is now known as saneax | 13:48 | |
| *** limao has joined #openstack-kuryr | 13:58 | |
| *** garyloug has joined #openstack-kuryr | 13:59 | |
| *** yedongcan_ has joined #openstack-kuryr | 14:05 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr: Add randomness to the returned vlan_ids https://review.openstack.org/422641 | 14:10 |
| *** hongbin has joined #openstack-kuryr | 14:20 | |
| *** yedongcan__ has joined #openstack-kuryr | 14:22 | |
| *** v1k0d3n has quit IRC | 14:23 | |
| *** yedongcan_ has quit IRC | 14:23 | |
| *** v1k0d3n has joined #openstack-kuryr | 14:24 | |
| *** yedongcan__ has quit IRC | 14:24 | |
| *** limao_ has joined #openstack-kuryr | 14:33 | |
| *** limao has quit IRC | 14:35 | |
| *** gsagie has quit IRC | 14:37 | |
| *** hongbin has quit IRC | 14:39 | |
| *** hongbin has joined #openstack-kuryr | 14:40 | |
| *** limao has joined #openstack-kuryr | 14:48 | |
| *** limao_ has quit IRC | 14:49 | |
| *** yedongcan__ has joined #openstack-kuryr | 14:49 | |
| *** yedongcan__ has quit IRC | 14:53 | |
| *** hongbin has quit IRC | 15:00 | |
| *** hongbin_ has joined #openstack-kuryr | 15:00 | |
| *** hongbin_ has quit IRC | 15:08 | |
| *** saneax is now known as saneax-_-|AFK | 15:12 | |
| openstackgerrit | Darla Ahlert proposed openstack/kuryr-libnetwork: Add reno support to kuryr-libnetwork https://review.openstack.org/424198 | 15:32 |
| *** janki has quit IRC | 15:33 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr: Add randomness to the returned vlan_ids https://review.openstack.org/422641 | 15:35 |
| *** hongbin has joined #openstack-kuryr | 15:56 | |
| *** limao has quit IRC | 15:59 | |
| *** reedip has quit IRC | 16:07 | |
| *** pcaruana has quit IRC | 16:13 | |
| *** reedip has joined #openstack-kuryr | 16:21 | |
| mchiappero | ltomasbo: have you run any benchmarks on nested containers with VLAN segmentation? | 16:34 |
| ltomasbo | mchiappero, I just create some containers and tested connectivity | 16:36 |
| ltomasbo | for both kuryr-libnetwork and kuryr-kubernetes | 16:36 |
| mchiappero | no, ok, I was just curious :) | 16:36 |
| mchiappero | do you plan to run some? | 16:36 |
| ltomasbo | anyway, I just tested on some devstack VM, so it would not be a fair benchmarking... | 16:37 |
| ltomasbo | maybe we can do so, yes. Do you have any specific benchmark in mind>? | 16:37 |
| mchiappero | uhm no, actually, but I was interested in comparing at some point VLAN, macvlan and ipvlan | 16:38 |
| mchiappero | maybe I'll consider allocating some time in the future if no one else will | 16:38 |
| mchiappero | well, it's more curiosity than a real interest | 16:39 |
| mchiappero | but could be useful | 16:39 |
| ltomasbo | were the ipvlan/macvlan port drivers merged? | 16:40 |
| mchiappero | on libnetwork only, we plan to start on k-k8s this week | 16:41 |
| ltomasbo | I meant for kubernetes | 16:41 |
| ltomasbo | great! | 16:41 |
| mchiappero | yeah, sorry, we've been stuck on other tasks and had little time for kuryr lately | 16:41 |
| ltomasbo | no problem, just curiosity! | 16:42 |
| mchiappero | same here :) | 16:42 |
| ltomasbo | would be really nice to try to use it with dpdk too | 16:42 |
| mchiappero | yes, we are considering it | 16:43 |
| ltomasbo | do you think it can work right away? making the VM runing the containers using the DPDK? Or will it be more complex than that? | 16:44 |
| mchiappero | running DPDK applications in containers is not a major issue, although for production use we still need a few bits that will likely be available in the coming weeks | 16:46 |
| mchiappero | integrating with COEs can be though, as it's a very different concept of networking | 16:46 |
| mchiappero | some orchestration engines only consider one (kernel backed) interface per container | 16:47 |
| mchiappero | I think I'll bring the topic up once ready | 16:48 |
| ltomasbo | ok, thanks mchiappero | 16:57 |
| *** david-lyle_ has joined #openstack-kuryr | 16:57 | |
| *** david-lyle_ has quit IRC | 17:03 | |
| mchiappero | ltomasbo: you're welcome! | 17:11 |
| *** devvesa has quit IRC | 17:28 | |
| *** david-lyle_ has joined #openstack-kuryr | 17:56 | |
| *** david-lyle_ has quit IRC | 17:56 | |
| *** david-lyle_ has joined #openstack-kuryr | 17:57 | |
| *** david-lyle_ is now known as david-lyle | 18:05 | |
| *** gsagie has joined #openstack-kuryr | 18:08 | |
| *** garyloug has quit IRC | 18:53 | |
| *** gsagie has quit IRC | 19:28 | |
| *** dougbtv_ has joined #openstack-kuryr | 20:22 | |
| *** dougbtv has quit IRC | 20:24 | |
| *** yuvalb has quit IRC | 21:09 | |
| *** oanson has quit IRC | 21:10 | |
| *** irenab has quit IRC | 21:10 | |
| *** yamamoto has joined #openstack-kuryr | 21:20 | |
| *** yuvalb has joined #openstack-kuryr | 21:29 | |
| *** oanson has joined #openstack-kuryr | 21:30 | |
| *** irenab has joined #openstack-kuryr | 21:30 | |
| *** yamamoto has quit IRC | 21:43 | |
| *** oanson has quit IRC | 21:56 | |
| *** oanson has joined #openstack-kuryr | 21:57 | |
| *** yamamoto has joined #openstack-kuryr | 22:23 | |
| *** oanson has quit IRC | 22:24 | |
| *** oanson has joined #openstack-kuryr | 22:25 | |
| *** limao has joined #openstack-kuryr | 22:46 | |
| *** saneax-_-|AFK is now known as saneax | 22:52 | |
| *** limao has quit IRC | 22:52 | |
| *** limao has joined #openstack-kuryr | 22:52 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!