A LoadBalancer service type only works in the cloud environment that supports external load balancers. This allows outside traffic to be routed into target Pods. In GCP, a TCP load balancer will be created by a LoadBalancer service type:

1. The firewall rules for allowing traffic between the load balancer and nodes will be created automatically:

// leveraging LoadBalancer service
# cat gke-service.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
spec:
  replicas: 1
  selector:
    matchLabels:
      run: nginx
  template:
    metadata:
      labels:
        run: nginx
    spec:
      containers:
        - image: nginx
          name: nginx
          ports:
            - containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  name: nginx
spec:
  ports:
    - port: 80
      targetPort: 80
  type: LoadBalancer
  selector:
    run: nginx

// create resources
# kubectl create -f gke-service.yaml
deployment "nginx" created
service "nginx" created

2. Let's check the service. The EXTERNAL-IP will show <pending> if the load balancer is still provisioning. Wait a while and the load balancer IP will present itself eventually:

# kubectl get svc nginx
NAME      TYPE           CLUSTER-IP      EXTERNAL-IP      PORT(S)        AGE
nginx     LoadBalancer   10.35.250.183   35.225.223.151   80:30383/TCP   11m

3. Let's curl $EXTERNAL-IP:80, to see if it works properly:

# curl -I 35.225.223.151
HTTP/1.1 200 OK
Server: nginx/1.13.9
Date: Thu, 01 Mar 2018 03:57:05 GMT
Content-Type: text/html
Content-Length: 612
Last-Modified: Tue, 20 Feb 2018 12:21:20 GMT
Connection: keep-alive
ETag: "5a8c12c0-264"
Accept-Ranges: bytes

4. If we check the forwarding rules in GCP, we can find a rule that defines how the traffic goes from external IP to the target pool:

# gcloud compute forwarding-rules list
NAME                              REGION       IP_ADDRESS      IP_PROTOCOL  TARGET
ae1f2ad0c1d0211e8858942010a80036  us-central1  35.225.223.151  TCP          us-central1/targetPools/ae1f2ad0c1d0211e8858942010a80036

5. A target pool is a set of instances that receive the traffic from forwarding rules. We could inspect the target pool by using the gcloud command as well:

// list target pools
# gcloud compute target-pools list
NAME                              REGION       SESSION_AFFINITY  BACKUP  HEALTH_CHECKS
ae1f2ad0c1d0211e8858942010a80036  us-central1  NONE                      k8s-1a4c86537c370d21-node

// check target pools info, replace $GCP_REGION as your default region.
# gcloud compute target-pools describe ae1f2ad0c1d0211e8858942010a80036 --region=$GCP_REGION
creationTimestamp: '2018-02-28T19:45:46.052-08:00'
description: '{"kubernetes.io/service-name":"default/nginx"}'
healthChecks:
- https://www.googleapis.com/compute/v1/projects/kubernetes-cookbook/global/httpHealthChecks/k8s-1a4c86537c370d21-node
id: '3515096241941432709'
instances:
- https://www.googleapis.com/compute/v1/projects/kubernetes-cookbook/zones/us-central1-a/instances/gke-my-k8s-cluster-default-pool-36121894-71wg
- https://www.googleapis.com/compute/v1/projects/kubernetes-cookbook/zones/us-central1-a/instances/gke-my-k8s-cluster-default-pool-36121894-04rv
- https://www.googleapis.com/compute/v1/projects/kubernetes-cookbook/zones/us-central1-a/instances/gke-my-k8s-cluster-default-pool-36121894-3mxm
kind: compute#targetPool
name: ae1f2ad0c1d0211e8858942010a80036
region: https://www.googleapis.com/compute/v1/projects/kubernetes-cookbook/regions/us-central1
selfLink: https://www.googleapis.com/compute/v1/projects/kubernetes-cookbook/regions/us-central1/targetPools/ae1f2ad0c1d0211e8858942010a80036
sessionAffinity: NONE

We can see there are three nodes inside the pool. Those are the same three nodes in our Kubernetes cluster. Load balancer will dispatch the traffic to a node based on a hash of the source/definition IP and port. A service with LoadBalancer type looks handy; however, it can't do path-based routing. It's time for ingress to come into play. Ingress supports virtual hosts, path-based routing, and TLS termination, which is a more flexible approach to your web services.