Bare metal configuration?

[thebigbone]

Hello. I was setting up caddy ingress controller on a bare metal server. I am aware that LoadBalancer IP would be in pending state so I added externalIPs to the Service type. After which I tried accessing using the allocated NodePort but it doesn't seem to be working. Any help would be appreciated! Thanks.

[axgkl]

I think (no expert here) k8s svc type loadbalancer (that's what this here creates) always requires a controller, which can, and will create/configure such an actual loadbalancer.

If you really have an external ip, there is metal lb which provides such a controller and which makes the external ip field of the caddy service go from pending to that real ip:

helm install metallb metallb/metallb -n metallb-system
k apply -f metallb_configmap_manifest.yaml 

with metallb_configmap_manifest:

apiVersion: v1
kind: ConfigMap
metadata:
  namespace: metallb-system
  name: config
data:
  config: |
    address-pools:
    - name: default
      protocol: layer2
      addresses:
      - 1.2.3.4/32 # <!------------- your ip

then check your caddy service - should now show a Loadbalancer with an external IP on it, just like in clouds.

Sorry if this was not what you asked for, pretty new within this world still 😊

[agusdallalba]

--namespace=caddy-system --set minikube=true --set replicaCount=1 is good enough for a 1-node cluster. You may delete the LoadBalancer-type Service.

You might have to give the caddy-system namespace the label pod-security.kubernetes.io/enforce: privileged.

It'd be cleaner to add 80,443 to the NodePort range or make caddy a DaemonSet.

With NodePort you can have 2 replicas for seamless upgrades. But traffic to a node without a caddy replica will do 2 hops to reach a Pod. With the DaemonSet you'd get one caddy per node in the cluster with no NAT, but no seamless upgrades.