I've got a single-node Kubernetes "cluster" built with kubeadm in AWS.

I have deployed a simple Nginx deployment with this config:

kind: Deployment
apiVersion: apps/v1
metadata:
  name: nginx0-deployment
  labels:
    app: nginx0-deployment
spec:
  replicas: 1
  selector:
    matchLabels:
      app: nginx0
  template:
    metadata:
      labels:
        app: nginx0
    spec:
      containers:
      - name: nginx
        image: k8s.gcr.io/nginx:latest
        ports:
        - containerPort: 80
          name: backend-http

I have also created an AWS ELB LoadBalancer:

kind: Service
apiVersion: v1
metadata:
  name: nginx0-service
balancing-enabled: "true"
spec:
  selector:
    app: nginx0-deployment
  ports:
  - protocol: TCP
    port: 80
    targetPort: 80
  type: LoadBalancer

This created the ELB and opened the relevant ports in the K8s instance security group.

{ec2-instance} ~ $ kubectl get all
NAME                                     READY   STATUS    RESTARTS   AGE
pod/nginx0-deployment-548f99f47c-ns75w   1/1     Running   0          3m45s

NAME                     TYPE           CLUSTER-IP       EXTERNAL-IP                                                                   PORT(S)        AGE
service/kubernetes       ClusterIP      10.96.0.1        <none>     443/TCP        25h
service/nginx0-service   LoadBalancer   10.106.179.191   acfc4150....elb.amazonaws.com   80:30675/TCP   63s

NAME                                READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/nginx0-deployment   1/1     1            1           3m45s

NAME                                           DESIRED   CURRENT   READY   AGE
replicaset.apps/nginx0-deployment-548f99f47c   1         1         1       3m45s

However something is still missing between the ELB and the POD because browsing to http://acfc4150....elb.amazonaws.com/ doesn't work - Chrome says ERR_EMPTY_RESPONSE.

I guess it's something to do with the ELB port mapping 80:30675/TCP - I have checked the incoming traffic on the instance and I see packets for port 30675 but nothing goes back. As if the mapping between this port and the POD's port 80 wasn't set up?

Any idea what should I add to my manifests to make it work?

Thanks!