example

Example usage of xds-controller with a server and client

gRPC client: time bounded caller that calls name service, No ports and no services are exposed. All it does it make a grpc call to server and log it. gRPC service: Name service that returns random name for every request

Generate proto

protoc -I=example/proto --go_out=example/proto example/proto/person.proto

protoc --go_out=. --go_opt=paths=source_relative \
    --go-grpc_out=. --go-grpc_opt=paths=source_relative \
    example/proto/person.proto

A Kubernates deployment of microservices

• client: polls for names continuously(every sec) for 3 minutes
• server: responds with random names

Run

To run example on a local cluster.,

Build

docker build -t xds.example/xds-controller .
docker build -t xds.example/client -f example/Dockerfile --build-arg TYPE=client example
docker build -t xds.example/server -f example/Dockerfile --build-arg TYPE=server example
docker pull envoyproxy/envoy:v1.24.0

Deployments

Pattern 1: CLient server communication using xds:look aside load balancing

kubectl create namespace xds-test
kubectl apply -f example/k8s/xds-controller.deployment.yaml -n xds-test
kubectl apply -f example/k8s/server.deployment.yaml -n xds-test
kubectl apply -f example/k8s/client.deployment.yaml -n xds-test
kubectl apply -f example/k8s/envoy.deployment.yaml -n xds-test

Pattern 2: Client with 2 servers communication using xds:look aside load balancing

kubectl create namespace xds-test
kubectl apply -f example/k8s/xds-controller.deployment.yaml -n xds-test
kubectl apply -f example/k8s/server.deployment.yaml -n xds-test
kubectl apply -f example/k8s/server.b.deployment.yaml -n xds-test
kubectl apply -f example/k8s/client.deployment.yaml -n xds-test

Pattern 3: External Client proxied via Envoy L7 Load balancer (No xds) using static dns

kubectl create namespace xds-test
kubectl apply -f example/k8s/server.deployment.yaml -n xds-test
kubectl apply -f example/k8s/envoy.deployment.yaml -n xds-test

Pattern 4: External Client proxied via Envoy L7 Load balancer. Evoy using xDS stream to discover server

kubectl create namespace xds-test
kubectl apply -f example/k8s/xds-controller.deployment.yaml -n xds-test
kubectl apply -f example/k8s/server.deployment.yaml -n xds-test
kubectl apply -f example/k8s/envoy.deployment.yaml -n xds-test

Metrics

xds-controller exposes metrics via /metrics endpoint by default on 8082 port. Portforwarding can be used to hit localhost and see the metric details

// get pods
kubectl get pods -n xds-test
kubectl port-forward {pod-name} 8082:8082 -n xds-test
curl localhost:8082/metrics

Test

Scale up and scaledown server pods and client should get notifications on changes and new pods should start getting the requests