Exposing Gloo with NodePort


By default, microservices deployed in Kubernetes have an internal flat network that is not accessible from the outside of the cluster. This is true even if you use Kubernetes on a public cloud (like Amazon AWS or Google Cloud).

A NodePort service is a way to make Kubernetes services available from outside the cluster (and potentially allow access from the internet) by opening ports on all of the nodes in the cluster and allowing traffic to go directly to the pods running within the cluster.

In this document, we will review how to expose Gloo via a NodePort service. It’s important to note that NodePort is not a recommended production setting. Node port has the following drawbacks not limited to:

  • You must coordinate what services use what ports so there is no conflict
  • A single service can be served on a port
  • You can only use ports 30000-32767
  • You can run into issues if your host/node/VMs change IP addresses.

See this article for more.

What is NodePort Service?

A Kubernetes cluster is composed of one or more nodes. A node VM is a (most likely) Linux machine (can be a virtual machine or bare-metal) that actually runs the Kubernetes pods.

When a Kubernetes service is created with NodePort type, Kubernetes chooses a port number and assigns it to the service. In addition, every node in the cluster is configured to forward traffic from this port to the pods belonging service.

This allows you to access the service simply by connecting to a node-ip:node-port where node-ip is the ip of any node in the cluster, and node-port is the NodePort assigned by Kubernetes.

One advantage of using a NodePort is that it allows relatively easy deployment on bare metal, as it does not depend on any load-balancing component outside the cluster.

How to use Gloo with NodePort?

In Gloo, the service that’s responsible for ingress traffic is called gateway-proxy. To use Gloo with NodePort we simply need to configure the gateway-proxy Kubernetes service to use NodePort. For example, when installing with Helm, use the following command:

helm install gloo/gloo --namespace gloo-system --set gatewayProxy.service.type=NodePort

Once installed, check what port was allocated:

kubectl get svc -n gloo-system gateway-proxy -o yaml

apiVersion: v1
kind: Service
    app: gloo
    gloo: gateway-proxy
  name: gateway-proxy
  namespace: gloo-system
  externalTrafficPolicy: Cluster
  - name: http
    nodePort: 30348
    port: 8080
    protocol: TCP
    targetPort: 8080
    gloo: gateway-proxy
  sessionAffinity: None
  type: NodePort
  loadBalancer: {}

In our example, port 30348 was allocated. You can now use http://NODE-IP:30348 to make requests to your Gloo virtual services.

If you are not using helm, or you already have Gloo installed, you can delete the existing Service, and expose a node port like this:

kubectl -n gloo-system delete svc gateway-proxy

kubectl -n gloo-system expose deploy/gateway-proxy \
  --name gateway-proxy \
  --type NodePort  \
  --port 80 \
  --target-port 8080

Note, if we manually expose it like this, we will only get a single port exposed. By default Gloo will use two ports, one for HTTP and one for HTTPS traffic. If you need to expose two ports, then you can do this directly in a yaml file.