API Platform Con 2021: join us in Lille (France) or online on September 10 for our 1st international conference.
API Platform Con 2021: join us in Lille (France) or online on September 10 for our 1st international conference.
API Platform Con 2021: join us in Lille (France) or online on September 10 for our 1st international conference.

Deploying to a Kubernetes Cluster

Kubernetes has become the most popular way to deploy, run and manage containers in production. Both Google Cloud Platform, Microsoft Azure and Amazon Web Services provide managed Kubernetes environment.

The official API Platform distribution contains a built-in Helm (the k8s package manager) chart to deploy in a wink on any of these platforms.

Preparing Your Cluster and Your Local Machine

  1. Create a Kubernetes cluster on your preferred Cloud provider or install Kubernetes locally on your servers
  2. Install Helm locally and on your cluster following their documentation
  3. Be sure to be connected to the right Kubernetes container e.g. running: gcloud config get-value core/project
  4. Update the Helm repo: helm repo update

Creating and Publishing the Docker Images

  1. Build the PHP and NGINX Docker images:

    docker build -t gcr.io/test-api-platform/php -t gcr.io/test-api-platform/php:latest api --target api_platform_php
    docker build -t gcr.io/test-api-platform/nginx -t gcr.io/test-api-platform/nginx:latest api --target api_platform_nginx
    docker build -t gcr.io/test-api-platform/varnish -t gcr.io/test-api-platform/varnish:latest api --target api_platform_varnish
  2. Push your images to your Docker registry, example with Google Container Registry:

    Docker client versions <= 18.03:

    gcloud docker -- push gcr.io/test-api-platform/php
    gcloud docker -- push gcr.io/test-api-platform/nginx
    gcloud docker -- push gcr.io/test-api-platform/varnish

    Docker client versions > 18.03:

    gcloud auth configure-docker
    docker push gcr.io/test-api-platform/php
    docker push gcr.io/test-api-platform/nginx
    docker push gcr.io/test-api-platform/varnish


Firstly you need to update helm dependencies by running:

helm dependency update ./api/helm/api

You are now ready to deploy the API!

Deploy your API to the container:

helm install ./api/helm/api --namespace=baz --name baz \
    --set php.repository=gcr.io/test-api-platform/php \
    --set nginx.repository=gcr.io/test-api-platform/nginx \
    --set secret=MyAppSecretKey \
    --set postgresql.postgresPassword=MyPgPassword \
    --set postgresql.persistence.enabled=true \
    --set corsAllowOrigin='^https?://[a-z\]*\.mywebsite.com$'

If you prefer to use a managed DBMS like Heroku Postgres or Google Cloud SQL (recommended):

helm install --name api ./api/helm/api \
    # ...
    --set postgresql.enabled=false \
    --set postgresql.url=pgsql://username:password@host/database?serverVersion=9.6

If you want to use a managed Varnish such as Fastly for the invalidation cache mechanism provided by API Platform:

helm install --name api ./api/helm/api \
    # ...
    --set varnish.enabled=false \
    --set varnish.url=https://myvarnish.com

Finally, build the client and admin JavaScript apps and deploy them on a static website hosting service.

Initializing the Database

Before running your application for the first time, be sure to create the database schema:

PHP_POD=$(kubectl --namespace=bar get pods -l app=php -o jsonpath="{.items[0].metadata.name}")
kubectl --namespace=bar exec -it $PHP_POD -- bin/console doctrine:schema:create

Tiller RBAC Issue

We noticed that some tiller RBAC trouble occurred. You can usually resolve it by running:

kubectl create serviceaccount --namespace kube-system tiller
  serviceaccount "tiller" created

kubectl create clusterrolebinding tiller-cluster-rule --clusterrole=cluster-admin --serviceaccount=kube-system:tiller
  clusterrolebinding "tiller-cluster-rule" created

kubectl patch deploy --namespace kube-system tiller-deploy -p '{"spec":{"template":{"spec":{"serviceAccount":"tiller"}}}}'
  deployment "tiller-deploy" patched

Please, see the related issue for further details / information. You can also take a look at the related documentation