GKE NodeJS App

This project illustrates how to build and deploy a simple app to the Google Kubernetes Engine and with support for canary deployments.

Quickstart

Setup the environment

Open the file ./scripts/set-env.sh and set the correct values for your environment.

./scripts/set-env.sh;

./scripts/gke-setup.sh;

./scripts/build-and-push-image.sh;
./scripts/deploy-version-production.sh;

export VERSION_NUMBER=2.0
./scripts/build-and-push-image.sh;
./scripts/deploy-version-canary.sh;

./scripts/gke-cleanup.sh;

The hard way

1. Setup GCP/Kubernetes

Setup the project:

gcloud config set project <PROJECT_ID>;
gcloud config set compute/zone <ZONE_ID>;

Create the cluster:

export CLUSTER_NAME=<CLUSTER_NAME>;
gcloud container clusters create $CLUSTER_NAME --num-nodes=3;
gcloud container clusters get-credentials $CLUSTER_NAME;

Create namespace:

kubectl create namespace production

2. Build and deploy production

Set version number:

export VERSION_NUMBER=1.0;

Build and push images to GCP:

export PROJECT_ID=$(gcloud info --format='value(config.project)');
docker build -t gcr.io/$PROJECT_ID/app:$VERSION_NUMBER .;
gcloud docker -- push gcr.io/$PROJECT_ID/app:$VERSION_NUMBER;

Deploy to GKE:

sed -i.bak "s/PROJECT_ID/$PROJECT_ID/; s/VERSION_NUMBER/$VERSION_NUMBER/;" k8s/app-production.yml;
kubectl --namespace=production apply -f k8s/app-production.yml
kubectl --namespace=production rollout status deployment/kubeapp-production

Make externally accessible:

kubectl --namespace=production apply -f k8s/app-ingress-production.yml
kubectl --namespace=production describe ingress
export SERVICE_IP=$(kubectl --namespace=production get ingress/app-ingress --output=json | jq -r '.status.loadBalancer.ingress[0].ip')
curl http://$SERVICE_IP/

4. Build and deploy canary

Set version number:

export VERSION_NUMBER=2.0;

Build and push images to GCP:

export PROJECT_ID=$(gcloud info --format='value(config.project)');
docker build -t gcr.io/$PROJECT_ID/app:$VERSION_NUMBER .;
gcloud docker -- push gcr.io/$PROJECT_ID/app:$VERSION_NUMBER;

Deploy to GKE:

sed -i.bak "s/PROJECT_ID/$PROJECT_ID/; s/VERSION_NUMBER/$VERSION_NUMBER/;" k8s/app-canary.yml;
kubectl --namespace=production apply -f k8s/app-canary.yml
kubectl --namespace=production rollout status deployment/kubeapp-canary

Create the ingress:

kubectl --namespace=production apply -f k8s/app-ingress-canary.yml
kubectl --namespace=production describe ingress
export INGRESS_IP=$(kubectl --namespace=production get ing/app-ingress --output=json | jq -r '.status.loadBalancer.ingress[0].ip')
echo "$INGRESS_IP foo.bar canary.foo.bar" | sudo tee -a /etc/hosts

5. Test HTTP Access

curl http://foo.bar/version
curl http://canary.foo.bar/version

6. Troubleshooting

kubectl --namespace=... get services;
kubectl --namespace=... get pods -o wide;
kubectl --namespace=... get events -w;

7. Cleanup

Delete the service:

kubectl --namespace=production delete service kubeapp-production-service
kubectl --namespace=production delete service kubeapp-canary-service

Wait for the Load Balancer provisioned for the app service to be deleted:

gcloud compute forwarding-rules list

Delete the container cluster:

gcloud container clusters delete $CLUSTER_NAME