NOTE: This has already been done, so no further action is needed, and the information below is only to show what has been done.
Due to port collision, port 10250 could not be used anymore for k3s metrics server, and had to be changed to port 4443.
This could be changed in the file /var/lib/rancher/k3s/server/manifests/metrics-server/metrics-server-deployment.yaml with commands
sudo sed -i 's/10250/4443/g' /var/lib/rancher/k3s/server/manifests/metrics-server/metrics-server-deployment.yaml
kubectl -n kube-system rollout status deployment metrics-server
but then the system would change this file back to using port 10250 again after k3s restart.
We have therefore applied a the following solution that allows for using a customized deployment file, while the
original deployment file is skipped.
kubectl -n kube-system get svc metrics-server \
-o jsonpath='{.spec.ports[0].port}→{.spec.ports[0].targetPort}{"\n"}'
# Expected: 443→https (or 443→4443 if you hard-coded it)
# Resolve the named targetPort:
kubectl -n kube-system get ep metrics-server \
-o jsonpath='{.subsets[0].ports[0].port}{"\n"}'
# → 4443
kubectl top nodes # must return CPU / memory numbers