How to restart kube-proxy?

[J1a-wei]

Hello RKE team,
We experienced a network outage in our production data center, which lasted for 30 minutes. Although the network is now back to normal, we have noticed that some kube-proxy instances have abnormal conditions, and we are unable to view the logs.

kubectl get po -o yaml kube-proxy-rke-worker-03
Error from server (NotFound): pods "kube-proxy-rke-worker-03" not found

I tried to delete it but couldn't, it remains in the Terminating state.
I plan to restart it, but I can't find a way.

Also, for the kube-proxy on worker-03, although it's in an abnormal state, I created an nginx pod and bound it to worker-03. Accessing its service still works.
So kube-proxy may still be working.

We are using the Cilium plugin with RKE2 version 1.26.
How should we proceed?

[brandond]

If the node is in an indeterminate state and containers are not manageable, I would probably recommend draining/cordoning it, followed by a reboot of the host. You could try just running the rke2-killall.sh script and then restarting rke2, but a reboot may be preferable.

For root cause, you might look at the kubelet and containerd logs around the time where the pod went into the terminating state.

[fabionitto]

what if the draining doesn't work?
Running rk2-killall.sh with bound pvc will remove data?

[brandond]

it'll unmount it. it should not delete data from the pvc.