I have deployed KubeFlow on an existing GKE cluster from here and want to create a LoadBalancer to access the UI. There is a ml-pipeline-ui-012 pod with an ml-pipeline-ui service having a type ClusterIP with a port 80/TCP. So I decided to expose the pod and create a service with a type LoadBalancer:

kubectl expose deployment ml-pipeline-ui --name ui-service --type=LoadBalancer --port 80 --target-port 80

After that the ui-service with a type Loadbalancer, an external IP and ports 80:30091/TCP was created. But I can't access it via https://external_ip:80 So I made a port forward for the pod:

kubectl port-forward ml-pipeline-ui-6d123-2x1 8080:30091

But this resulted in an error:

an error occurred forwarding 8080 -> 80: error forwarding port 80 to pod e678f12345, uid : exit status 1: 2020/03/05 09:50:17 socat[23031] E connect(5, AF=2 127.0.0.1:80, 16): Connection refused 

What am I doing wrong here?