Kubernetes - Services

Prerequisites

• Kubernetes
• kubectl

Assignment

• Lets start by creating a deployment and exposing it as a service using different options.

Deployment

kubectl create deployment nginx --image=nginx --port=80 --replicas=3 -o yaml --dry-run=client

kubectl create deployment nginx --image=nginx --port=80 --replicas=3

ClusterIP Service

1️⃣ Create a simple deployment using nginx and expose it on port 80 with 3 replicas.

kubectl expose deployment/nginx --dry-run=client -o yaml

kubectl expose deployment/nginx

2️⃣ Check the exposed services and endpoints on the kubernetes cluster. Note that all the endpoints relate to the nginx pods ip addresses.

kubectl get services

kubectl get endpoints

kubectl get pods -o wide

3️⃣ Capture the Cluster IP and access the exposed service using curl.

CLUSTER_IP=$(kubectl get services | grep ClusterIP | grep nginx | awk {'print $3'}); echo $CLUSTER_IP

curl $CLUSTER_IP

4️⃣ Try accessing the service from another pod and check the DNS search path relating to the service.

kubectl run --rm -it curl --image=curlimages/curl:8.4.0 --restart=Never -- sh

curl nginx.default.svc.cluster.local

cat /etc/resolv.conf

curl nginx

exit

5️⃣ Delete the service.

kubectl delete service/nginx

NodePort Service

1️⃣ Expose the service using NodePort and verify that two ports are listed - Application and NodePort.

kubectl expose deployment/nginx --type=NodePort

kubectl get service

2️⃣ Capture the Cluster IP and the NodePort to query the NodePort service from one of the nodes.

kubectl get nodes -o wide

CONTROL_PLANE_IP=$(kubectl get nodes -o wide | grep control-plane | awk {'print $6'}); echo $CONTROL_PLANE_IP

NODEPORT_PORT=$(kubectl get services | grep NodePort | grep nginx | awk -F'[:/]' '{print $2}'); echo $NODEPORT_PORT

curl ${CONTROL_PLANE_IP}:${NODEPORT_PORT}

3️⃣ Delete the NodePort service.

kubectl delete service/nginx

LoadBalancer Service

1️⃣ Expose the deployment using the LoadBalancer service type.

kubectl expose deployment/nginx --type=LoadBalancer --port 8080 --target-port 80

kubectl get service

2️⃣ Capture the IP address and the port and query the nginx service to check how the load balancer routes the query to different pods.

LOADBALANCER_IP=$(kubectl get service | grep LoadBalancer | grep nginx | awk '{split($0,a," "); split(a[4],b,","); print b[1]}'); echo $LOADBALANCER_IP

LOADBALANCER_PORT=$(kubectl get service | grep LoadBalancer | grep nginx | awk -F'[:/]' '{print $2}'); echo $LOADBALANCER_PORT

watch --differences "curl ${LOADBALANCER_IP}:${LOADBALANCER_PORT} 2>/dev/null"

3️⃣ Delete the LoadBalancer service

kubectl delete deployment/nginx service/nginx

ExternalName Service

1️⃣ Create two deployments with different images.

kubectl create deployment nginx-red --image=spurin/nginx-red --port=80

kubectl create deployment nginx-blue --image=spurin/nginx-blue --port=80

kubectl get deployment

2️⃣ Expose both the deployments as a service.

kubectl expose deployment/nginx-red

kubectl expose deployment/nginx-red

kubectl get service

3️⃣ Create an ExternalName service of my-service that points to nginx-red pod.

kubectl create service externalname my-service --external-name nginx-red.default.svc.cluster.local

kubectl get service

4️⃣ Spin up a new pod and curl all the services. Notice that the ExternalName service points to nginx-red pod.

kubectl run --rm -it curl --image=curlimages/curl:8.4.0 --restart=Never -- sh

curl nginx-red

curl nginx-blue

curl my-service

nslookup my-service

5️⃣ In another tab, edit the server my-service and update it to point to nginx-blue.

kubectl edit service/my-service

6️⃣ In the original tab where we have ssh’ed into the curl container, check the my-service resolution.

nslookup my-service

curl my-service

exit

7️⃣ Cleanup the deployment and service for ExternalName service.

kubectl delete deployment/nginx-blue deployment/nginx-red service/nginx-blue service/nginx-red service/my-service

Headless Service

1️⃣ Create a nignx deployment with 3 replicas.

kubectl create deployment nginx --image=spurin/nginx-debug --replicas=3 --port=80

2️⃣ Create a headless service by modifying the ClusterIP yaml declaration and apply the changes.

kubectl expose deployment/nginx --dry-run=client -o yaml --type=ClusterIP | tee headless.yaml

grep -q 'clusterIP: None' headless.yaml || sed -i '/spec:/a\ \ clusterIP: None' headless.yaml; cat headless.yaml

kubectl apply -f headless.yaml 

3️⃣ Verify if the headless service was created successfully.

kubectl get service

4️⃣ Run the curl pod and note the IP address that are reference while we curl the headless service.

kubectl run --rm -it curl --image=curlimages/curl:8.4.0 --restart=Never -- sh

watch nslookup nginx

curl nginx

exit

5️⃣ Remove the headless service.

rm headless.yaml; kubectl delete deployment/nginx service/nginx