Multicluster Gateways With Kubernetes Gateway API

Kubernetes Gateway API is the new specification released by CNCF to standardize the Kubernetes Ingress traffic. Now, what if a service is configured as High Availability (HA)? (Say it is in a different cloud environment and you have to access it from the Gateway; i.e., multicluster, multi-cloud scenario.) In this article, we will showcase how to use the Gateway API spec to configure gateways for multicluster setup. 

Multicluster Kubernetes Gateway Demo Overview

We have two clusters: one in EKS (primary) and the other in GKE (remote). I have deployed Istio in both the clusters and the setup is primary-remote Istio installation. Istio is used as the controller to implement the Gateway API resources. 

Here’s what I’m going to do:

• In the primary cluster/EKS, deploy the helloworld-v1 deployment, helloworld service, and echoserver service.
• In the remote cluster/GKE, deploy helloworld-v2 deployment, helloworld service, echoserver deployment, and echoserver service.
• Deploy the Kubernetes Gateway API resources — Gateway and HTTPRoutes — in the primary cluster.
• After the deployments, we will verify that the Gateway in the primary cluster/EKS can access the services in the remote cluster/GKE, as shown in the image below:

Multicluster, multi-cloud Gateway with K8s Gateway API demo setup

Deploy the Applications and Services in Clusters

Deploy helloworld-service in both the primary and remote clusters:

kubectl -f apply helloworld-service.yaml --context=eks-cluster
kubectl -f apply helloworld-service.yaml --context=gke-cluster

kubectl -f apply helloworld-deployment-v1.yaml --context=eks-cluster
kubectl -f apply helloworld-deployment-v2.yaml --context=gke-cluster

kubectl -f apply echoserver-service.yaml --context=eks-cluster
kubectl -f apply echoserver-service.yaml --context=gke-cluster
kubectl -f apply echoserver-deployment.yaml --context=gke-cluster

Note that service resources need to be deployed in both clusters for this to work. That is why I deployed the echoserver-service in the primary cluster/EKS although the deployment is only in the remote cluster/GKE.

Now, let us verify the deployments in both the primary and secondary clusters:

kubectl get svc -n demo --context=eks-cluster
kubectl get pods -n demo --context=eks-cluster
kubectl get svc -n demo --context=gke-cluster
kubectl get pods -n demo --context=gke-cluster

The primary cluster has the helloworld-v1 pod running, while the remote cluster has both helloworld-v2 and echoserver pods running successfully:

Deploy K8s Gateway API Resources and Verify Multicluster Communication

Apply the gateway resource in the primary/EKS cluster:

kubectl apply -f gateway-api-gateway.yaml --context=eks-cluster

The Gateway uses Istio as the controller and is deployed in the istio-ingress namespace.

Deploy HTTPRoute in the primary cluster for the helloworld application, which listens on path /hello:

kubectl apply -f helloworld-httproute.yaml --context=eks-cluster

Now, let us verify multicluster communication by curling the helloworld application; but first, we need to get the Gateway IP:

kubectl get svc -n istio-ingress --context=eks-cluster

curl your_gateway_external_ip/hello

You can see that the request is served by both the helloworld-v1 and helloworld-v2 that are deployed in the primary and secondary clusters, respectively.

Now, let us deploy the HTTPRoute for echoserver in the primary cluster, which listens on / :

kubectl apply -f echoserver-httproute.yaml --context=eks-cluster

Verify if the Gateway is able to access echoserver deployed in the remote cluster:

curl your_gateway_external_ip

The Gateway is able to get a response from echoserver deployed in the remote cluster successfully. And that is the end of the demo.