Add instructions how to configure traefik

[aDingil]

Hey @mysticaltech,

i´m currently want to add a middleware to traefik, but i´m not sure about the how.
My current tries destroy the given traefik config from terraform.

I can create prs to improve the examples/instructions after understanding the solution 😄

Regards

[aDingil]

nvm about the middleware, solved it myself.
But how can i expose the traefik dashboard?

[mysticaltech]

@aDingil Try searching the traefik or k3s repo themselves, or even ask GPT-4. I wish I knew, but it's probably super simple.

[aDingil]

Hey,

thx for your quick response and sry for my delay....
So i ask GPT 3.x/4, visited the k3s docs and other things, but still no success.
I wasn´t able to port forward to my localhost or expose the dashboard to a hostname.

Some observations:

• traefik is not running in the namespace kube-system, it is running in the traefik namespace
• this is the only recent guide i was able find
• There a numerous issues about this topic in the k3s project

It is super weird, that this simple task is so obfuscated without any "official" docs from k3s.
Also for people who using helm there is the concept about helmchartconfig to modify packaged components like traefik

I just hope that i wouldn´t be dependent on the traefik dashboard to solve/debug a problem in the future.

Regards

[aDingil]

I don´t how and why, but now it does work:

List of changes:

• Created ingress and service yamls, described here, but changed the namespace to traefik
• Created a helm config

apiVersion: helm.cattle.io/v1
kind: HelmChartConfig
metadata:
  name: traefik
  namespace: traefik
spec:
  valuesContent: |-
    additionalArguments:
      - "--api"
      - "--api.dashboard=true"
      - "--api.insecure=true"
      - "--log.level=DEBUG"
    ports:
      traefik:
        expose: true
    providers:
      kubernetesCRD:
        allowCrossNamespace: true

• Created a ingressroute for traefik

apiVersion: traefik.containo.us/v1alpha1
kind: IngressRoute
metadata:
  annotations:
    meta.helm.sh/release-name: traefik
    meta.helm.sh/release-namespace: traefik
  generation: 8
  labels:
    app.kubernetes.io/instance: traefik-traefik
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/name: traefik
  name: traefik-dashboard
  namespace: traefik
spec:
  entryPoints:
  - web
  routes:
  - kind: Rule
    match: Host(`HOST_URL`)
    services:
    - kind: TraefikService
      name: api@internal

Eventlog

Podlist

I also tried to reverted all those changes, to see when it stops the dashboard exposing, but it is still running and i still don´t how to stop or modify it

[mysticaltech]

@aDingil This seems activated by your custom helmchartconfig. Just kubectl delete it if you want to stop that. However, you do not need the full helmchartconfig in your case, just use the traefik_additional_options variable (see kube.tf.example), in there you can pass all options needed for the dashboard. Basically, those:

["--api", "--api.dashboard=true", "--api.insecure=true"]

[mysticaltech]

Or even better, just pass your own traefik_values via the variable! Again see kube.tf.example. That would render all other traefik variables meaningless, and allow for more customization.

Here's where the logic lives if curious https://github.com/kube-hetzner/terraform-hcloud-kube-hetzner/blob/master/locals.tf#L421.

[aDingil]

Many thx for the additional info.

I removed the helmchartconfig:

I don´t even have an ingress(route) defined anymore in my chart

But the dashboard is still exposed, maybe it´s a dns/browser cache thing dunno.

[mysticaltech]

No way the ingress definition has been deleted and the dashboard is still showing! Please run kubectl get ing -A.

[aDingil]

all ingresses

kubectl get ing -A  --kubeconfig xxx
NAMESPACE   NAME              CLASS     HOSTS                            ADDRESS   PORTS     AGE
default     mailhog-ingress   traefik   mailhog                                    80, 443   7d2h

all ingressroutes

kubectl get ingressroutes --all-namespaces --kubeconfig xxx
No resources found

[mysticaltech]

Then check where your DNS is pointing to, with dig a or dig cname, because the dashboard is not coming from the cluster. Unless you've setup some sort of port forward.

[aDingil]

you are absolutely right, i was using a different domain and thought it was the kubernetes domain.
upsi

[aDingil]

i will postpone the dashboard stuff, because i guess i will not needed, fingers crossed.
Thx again for your hard work and help.