| *** pcaruana has joined #openstack-kuryr | 05:24 | |
| *** pcaruana has quit IRC | 05:30 | |
| *** janki has joined #openstack-kuryr | 05:35 | |
| *** gdwornicki has quit IRC | 06:36 | |
| *** pcaruana has joined #openstack-kuryr | 06:41 | |
| *** ccamposr has joined #openstack-kuryr | 07:19 | |
| *** jchhatbar has joined #openstack-kuryr | 07:23 | |
| *** janki has quit IRC | 07:25 | |
| *** gdwornicki has joined #openstack-kuryr | 08:00 | |
| openstackgerrit | Maysa de Macedo Souza proposed openstack/kuryr-kubernetes master: Ensure no LBaaS SG update is triggered for SVCs without selectors and ports https://review.opendev.org/685626 | 08:08 |
|---|---|---|
| openstackgerrit | Merged openstack/kuryr-kubernetes master: Ensure pod IP is retrieved from annotation https://review.opendev.org/685114 | 08:57 |
| *** mrostecki has quit IRC | 09:39 | |
| *** mrostecki has joined #openstack-kuryr | 09:41 | |
| openstackgerrit | Maysa de Macedo Souza proposed openstack/kuryr-kubernetes master: Ensure no LBaaS SG update is triggered for SVCs without selectors and ports https://review.opendev.org/685626 | 09:41 |
| *** gcheresh_ has joined #openstack-kuryr | 09:55 | |
| dulek | gdwornicki: What do you mean by "bad port"? | 09:56 |
| dulek | gdwornicki: Any logs we can take a look at? | 09:56 |
| gdwornicki | I will send them to you next time user will deploy. I will be a matter of time today. They deploy from helm like ~10 pod and services and never they had a clean deployment. | 09:59 |
| gdwornicki | However if you do I in steps like 2-4 smaller deployment it works | 10:00 |
| gdwornicki | Usually error is ResourceNotReady and it is not catched | 10:01 |
| gdwornicki | Other problem I can point out is that users tent to forgot about the address space and they try to deploy more ports than address space of the ns. If they would get error it would be OK. But kontroler fails and is not serving new requests | 10:02 |
| gdwornicki | This second is easy to reproduce | 10:02 |
| gdwornicki | Just create two ns with namespace driver. Fill one out of the address space and try to deploy something on the good one | 10:03 |
| gdwornicki | Sry for my spelling wring from phone in hurry | 10:04 |
| dulek | gdwornicki: Are you using nested vif driver? The first issue sounds a bit like the problem with Neutron subports that we faced a while ago. | 10:12 |
| dulek | Lemme try to find the bugreport. | 10:12 |
| gdwornicki | Yes we are using vlan nested driver | 10:12 |
| dulek | https://bugs.launchpad.net/neutron/+bug/1828375 - this thing. | 10:13 |
| openstack | Launchpad bug 1828375 in neutron "Bulk creation of subports fails with StaleDataError" [Low,In progress] - Assigned to Bence Romsics (bence-romsics) | 10:13 |
| gdwornicki | And if I may ask. Will it work with vxlan networks? We are going to migrate. | 10:14 |
| dulek | gdwornicki: Please check if this could be it. | 10:14 |
| gdwornicki | Ok | 10:14 |
| dulek | gdwornicki: And about the second issue - what would be the expected behavior on subnet exhaustion? | 10:14 |
| gdwornicki | Yes in case of first ns | 10:15 |
| gdwornicki | That what is filled up | 10:15 |
| gdwornicki | But second should work as it has free addresses. But the controller throws exceptions and crashes | 10:16 |
| dulek | gdwornicki: Ah, I get it, it affects a "good" namespace as well. Can you file a bug then? We'll take a look. | 10:16 |
| gdwornicki | Making cni useless - each restart fails due to the filled up ns. The operator needs to clean up the bad ns to make cluster useful | 10:16 |
| gdwornicki | Ok. I have logs on it I shall post it later on leave here bug number ok? | 10:17 |
| gdwornicki | I don't have access to my computer now | 10:17 |
| dulek | gdwornicki: Sure, please ping me once it's filed. | 10:27 |
| *** jchhatbar has quit IRC | 11:00 | |
| *** gdwornicki has quit IRC | 11:19 | |
| *** gcheresh_ has quit IRC | 12:40 | |
| *** gcheresh_ has joined #openstack-kuryr | 12:50 | |
| *** aperevalov has joined #openstack-kuryr | 13:06 | |
| aperevalov | dulek, ltomasbo hi, could we make a sync (meeting today) at 3 p.m. UTC today? | 13:14 |
| dulek | aperevalov: Sure! | 13:15 |
| aperevalov | here in this channel? | 13:15 |
| dulek | aperevalov: Yeah, that would be best. | 13:20 |
| *** gcheresh_ has quit IRC | 13:47 | |
| openstackgerrit | Merged openstack/kuryr-kubernetes master: Ensure no LBaaS SG update is triggered for SVCs without selectors and ports https://review.opendev.org/685626 | 13:55 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes stable/train: Ensure pod IP is retrieved from annotation https://review.opendev.org/685718 | 14:05 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes stable/train: Ensure no LBaaS SG update is triggered for SVCs without selectors and ports https://review.opendev.org/685719 | 14:05 |
| *** pcaruana has quit IRC | 14:18 | |
| *** brault has joined #openstack-kuryr | 14:50 | |
| *** gdwornicki has joined #openstack-kuryr | 14:54 | |
| aperevalov | I was not able to participate in #openstack-meeting-4 at 2p.m. UTC. | 14:56 |
| dulek | aperevalov: Hello! So what's up? | 14:57 |
| dulek | Pinging ltomasbo as well. | 14:57 |
| aperevalov | about optimization task, I told about implementation of direct RPC instead of using k8s, but I measured again and found neither kubernetes is a bottle neck nor kuryr-kuberntes, just my previous measurement was incorrect. | 14:59 |
| dulek | aperevalov: So what's the bottleneck? | 15:00 |
| aperevalov | dulek, latest measurement shows there is no bottleneck here. Regarding improvement in pool (precreated net namespace), we don't still use port pool, so it's not our use case, and we postponed this task. | 15:01 |
| dulek | aperevalov: Okay! That's rather a good news, right? | 15:01 |
| aperevalov | yes ;) next topic is SR-IOV. Now we a going to move to sriovdp 3.0. Right before we used 2.1.0 and 2.0.0. As I remember Danil did not mention in doc (required version). BUT | 15:02 |
| dulek | aperevalov: sriovdp is a library? | 15:03 |
| aperevalov | it implemented as k8s device-plugin, and runs in the pod in kube-system namespace. But it also implement k8s hooks to add env variable and mount points into user pod. kuryr-kubernetes binding driver's behaviour depends on sriovdp version, especially in part of vfio driver. Danil now in vacation, but I hope when he will return back we'll make it | 15:05 |
| aperevalov | up. | 15:05 |
| dulek | Okay, very well. | 15:06 |
| dulek | aperevalov: You're probably aware that Train release is close now. We might not be able to get it into this release. | 15:06 |
| dulek | aperevalov: Does it matter for you? At RH we're already maintaining a fork that we release along with OpenShift releases. | 15:07 |
| dulek | So we don't really care about OpenStack releases that much anymore. | 15:07 |
| aperevalov | Yes, I know, about code freeze. I don't think it's important for us.Integration tests for SRIOV: you are using tempest framework, but our CI/CD team've preferred sonobuoy. I already implemented tempest, but need to finish it. So I'll publish it soon. | 15:07 |
| aperevalov | fork of kuryr-kubernetes? | 15:08 |
| dulek | aperevalov: Okay, very well. Also please note that we're not constrained to tempest in the gate and had already experimented with running Kubernetes upstream tests. | 15:09 |
| dulek | aperevalov: Yup, there: https://github.com/openshift/kuryr-kubernetes | 15:09 |
| dulek | But don't worry, we still develop on OpenStack, that fork is only required for procedural reasons - OpenShift requires code to live on github.com/openshift. | 15:10 |
| aperevalov | thanks for link ) | 15:10 |
| dulek | So we only rebase it with new code. | 15:10 |
| aperevalov | we also have a fork of kuryr-kuberntes, and the main difference, from my point of view in multi vif driver, we don't use NPWG driver for it, we're using our own driver. | 15:11 |
| dulek | aperevalov: Okay, so you don't really care about OpenStack releases either. This is good. | 15:12 |
| aperevalov | Our collogue from Suwon wanted to visit Open Infrastructure summit to discuss the reason we're doing it in our own way and possibilities to upstream it ) | 15:12 |
| aperevalov | also last time (before latest summit) Daniel asked me about sriov works. | 15:13 |
| dulek | aperevalov: Ah, I've seen the email from Yash. I replied, but haven't heard from him yet. | 15:13 |
| aperevalov | ok, this time I'm 100% sure I'll wil not attend at summit. But for the next summit I hope to prepare presentation about optimized network in containers managed by neutron. | 15:15 |
| dulek | aperevalov: I would be really willing to hear about that. :) | 15:16 |
| aperevalov | great, that's all what I wanted to say. | 15:18 |
| dulek | Thank you! | 15:18 |
| aperevalov | thank you too! | 15:19 |
| openstackgerrit | Sarka Scavnicka proposed openstack/kuryr-kubernetes master: [WIP] Add a loadbalancer CRD https://review.opendev.org/685737 | 15:44 |
| *** pcaruana has joined #openstack-kuryr | 15:46 | |
| *** gdwornicki has quit IRC | 16:08 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Ensure lb sg rules are updated upon namespace label updates https://review.opendev.org/685754 | 16:23 |
| *** pcaruana has quit IRC | 20:50 | |
| *** irclogbot_1 has quit IRC | 20:54 | |
| *** irclogbot_3 has joined #openstack-kuryr | 20:56 | |
| *** irclogbot_3 has quit IRC | 21:01 | |
| *** irclogbot_0 has joined #openstack-kuryr | 21:02 | |
| *** irclogbot_0 has quit IRC | 21:07 | |
| *** irclogbot_1 has joined #openstack-kuryr | 21:08 | |
| *** irclogbot_1 has quit IRC | 21:13 | |
| *** irclogbot_0 has joined #openstack-kuryr | 21:14 | |
| *** irclogbot_0 has quit IRC | 21:19 | |
| *** irclogbot_3 has joined #openstack-kuryr | 21:20 | |
| *** irclogbot_3 has quit IRC | 21:25 | |
| *** irclogbot_3 has joined #openstack-kuryr | 21:26 | |
| *** irclogbot_3 has quit IRC | 21:33 | |
| *** irclogbot_0 has joined #openstack-kuryr | 21:36 | |
| *** irclogbot_0 has quit IRC | 21:41 | |
| *** irclogbot_0 has joined #openstack-kuryr | 21:42 | |
| *** irclogbot_0 has quit IRC | 21:47 | |
| *** irclogbot_0 has joined #openstack-kuryr | 21:48 | |
| *** irclogbot_0 has quit IRC | 21:53 | |
| *** irclogbot_2 has joined #openstack-kuryr | 21:54 | |
| *** irclogbot_2 has quit IRC | 21:59 | |
| *** irclogbot_2 has joined #openstack-kuryr | 22:00 | |
| *** irclogbot_2 has quit IRC | 22:05 | |
| *** irclogbot_1 has joined #openstack-kuryr | 22:06 | |
| *** irclogbot_1 has quit IRC | 22:15 | |
| *** irclogbot_1 has joined #openstack-kuryr | 22:16 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!