*** ccamposr has joined #openstack-kuryr | 00:10 | |
*** rh-jelabarre has quit IRC | 03:10 | |
*** ndesh has joined #openstack-kuryr | 06:43 | |
*** gcheresh has joined #openstack-kuryr | 07:06 | |
*** ltomasbo has joined #openstack-kuryr | 07:22 | |
*** gcheresh has quit IRC | 07:56 | |
*** maciejjozefczyk has joined #openstack-kuryr | 07:59 | |
*** maciejjozefczyk_ has joined #openstack-kuryr | 08:04 | |
*** maciejjozefczyk has quit IRC | 08:05 | |
*** gcheresh has joined #openstack-kuryr | 08:09 | |
ndesh | dulek I was working kubernetes jobs when I noticed this behavior. When the job is done and hence the pod is in the completed state, the kubelet issues a CNI DEL request which leads to the removal of the network interface from the pod. After this the corresponding neutron port becomes DOWN. In the current code we don't delete the neutron port until | 08:13 |
---|---|---|
ndesh | the job itself is deleted. But I think we should delete the neutron port once the pod is in the completed state and remove the corresponding annotation as we have received the CNI DEL request. What are you thoughts on this? | 08:13 |
dulek | ndesh: Oh, interesting case. | 08:17 |
dulek | ndesh: Are we guaranteed that pod will never get restarted? Probably yes, but it'd be good to confirm that. | 08:18 |
dulek | ndesh: Then kuryr-daemon could mark the pod with an annotation that kuryr-controller would notice and process. | 08:19 |
dulek | That would work, wouldn't it? | 08:19 |
dulek | Alternatively kuryr-controller could just notice pods in state "Completed" and remove ports for those, but then we might have race condition between kuryr-daemon and kuryr-controller - maybe we shouldn't remove port before kuryr-daemon unplugs it. | 08:20 |
dulek | Not really sure about last one. | 08:20 |
ndesh | Yes we need to see the race conditions here | 08:21 |
ndesh | I will try to research more on the issue try to create a patch to handle it | 08:22 |
*** maciejjozefczyk_ is now known as maciejjozefczyk | 09:01 | |
*** gcheresh has quit IRC | 10:21 | |
*** openstackgerrit has joined #openstack-kuryr | 10:28 | |
openstackgerrit | Merged openstack/kuryr-tempest-plugin master: Retry exec_command_in_pod https://review.opendev.org/712982 | 10:28 |
*** gcheresh has joined #openstack-kuryr | 10:30 | |
*** gcheresh has quit IRC | 10:41 | |
openstackgerrit | Michał Dulko proposed openstack/kuryr-kubernetes master: Preserve unknown in remote_ip_prefixes in NP CRD https://review.opendev.org/713625 | 10:55 |
openstackgerrit | Michał Dulko proposed openstack/kuryr-kubernetes master: Preserve unknown in remote_ip_prefixes in NP CRD https://review.opendev.org/713625 | 10:58 |
*** gcheresh has joined #openstack-kuryr | 11:03 | |
openstackgerrit | Merged openstack/kuryr-kubernetes master: do not die if cannot remove socket file https://review.opendev.org/711463 | 11:13 |
*** gcheresh has quit IRC | 11:43 | |
*** rh-jelabarre has joined #openstack-kuryr | 11:53 | |
*** gcheresh has joined #openstack-kuryr | 11:55 | |
openstackgerrit | Merged openstack/kuryr-kubernetes master: Autogenerate path for vhostuserclient socket https://review.opendev.org/711464 | 13:20 |
*** ltomasbo has quit IRC | 14:43 | |
*** luis5tb has joined #openstack-kuryr | 14:54 | |
openstackgerrit | Merged openstack/kuryr-kubernetes master: Add IPv6 support to namespace subnet driver https://review.opendev.org/708657 | 15:53 |
*** luis5tb has quit IRC | 17:17 | |
*** gcheresh has quit IRC | 18:43 | |
*** maciejjozefczyk has quit IRC | 18:56 | |
*** gcheresh has joined #openstack-kuryr | 19:06 | |
*** ndesh has quit IRC | 19:45 | |
*** gcheresh has quit IRC | 19:54 | |
*** gmann is now known as gmann_afk | 19:57 | |
*** gmann_afk is now known as gmann | 20:33 | |
*** gcheresh has joined #openstack-kuryr | 20:39 | |
*** gcheresh has quit IRC | 21:12 | |
*** threestrands has joined #openstack-kuryr | 22:58 | |
*** maysams has quit IRC | 23:36 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!