Unable to Disable Traefik LoadBalancer in K3s

[vkoori]

I am attempting to disable the Traefik load balancer when starting my K3s server, but the Traefik service is still being created as a LoadBalancer type. I used the following command to start K3s with the required flags:

ExecStart=/usr/local/bin/k3s \
    server \
    --disable=traefik \
    --disable=servicelb \
    --disable=local-storage \
    '--write-kubeconfig-mode' \
    '644'

However, after starting K3s, I can still see the Traefik service running as a LoadBalancer in the kube-system namespace when I query the services with kubectl get service -n kube-system.

Output:

$ kubectl get service -n kube-system
NAME             TYPE           CLUSTER-IP      EXTERNAL-IP     PORT(S)                      AGE
kube-dns         ClusterIP      10.43.0.10      <none>          53/UDP,53/TCP,9153/TCP       101m
metrics-server   ClusterIP      10.43.89.231    <none>          443/TCP                      2d1h
traefik          LoadBalancer   10.43.139.147   192.168.10.10   80:30639/TCP,443:30764/TCP   15h

Expected Behavior:
I expect the Traefik load balancer to be completely disabled and not be present as a LoadBalancer service in the kube-system namespace.

[brandond]

Can you show the output of kubectl get node -o yaml | grep node-args and kubectl get pod -A -o wide?

[vkoori]

Thanks for your response!

$ kubectl get node -o yaml | grep node-args
      k3s.io/node-args: '["server","--disable","traefik","--disable","servicelb","--disable","local-storage","--write-kubeconfig-mode","644"]'

$ kubectl get pod -A -o wide
NAMESPACE     NAME                               READY   STATUS    RESTARTS        AGE   IP           NODE     NOMINATED NODE   READINESS GATES
kong          kong-controller-6845694c68-gplbq   1/1     Running   17 (120m ago)   39h   10.42.0.68   kooriv   <none>           <none>
kong          kong-gateway-685b68bc4c-k44j8      1/1     Running   1 (120m ago)    21h   10.42.0.65   kooriv   <none>           <none>
kube-system   coredns-ccb96694c-h9j4b            1/1     Running   1 (120m ago)    20h   10.42.0.66   kooriv   <none>           <none>
kube-system   metrics-server-5985cbc9d7-njgkq    1/1     Running   1 (120m ago)    20h   10.42.0.67   kooriv   <none>           <none>

However, my issue isn't related to the node args or any pods. The problem I'm facing is specifically with the LoadBalancer service for Traefik that persists even after starting K3s with the --disable=traefik and --disable=servicelb flags.

[brandond]

The service should go away when traefik is disabled and the chart uninstalled. I've never seen it leave anything behind. I don't see an uninstall job pod present - did it succeed, or did you delete anything manually?

[vkoori]

nope. I dont delete anything manually

[luo3house]

The docs indicates --disable=traefik, but I turned to --disable traefik then it worked.

[brandond]

The flag parser should handle either.