HaElixir

An example of how to deploy multiple instances of a single elixir release while setting an environment variable differently for each instance. This is useful in the case of breaking up a monolithic applications into stateless and stateful parts that can be ran independently.

This example is intended to be used with minikube and helm.

Initialization

Run make init to setup the helm repository and start minikube. This also tells docker to reuse the minikube docker daemon. To go back to a state where that is not the case, run make unset-docker.

Build

Run make build to build the docker image to be used in minikube

Deploy

Run make deploy to deploy into minikube. This will deploy postgres, run database migrations and then start up 4 deployments, each with a different SERVICE environment variable set. This environment variable is then displayed in the greeting message at the root path of each deployment.

Delete

Run make delete to delete all deployed infrastructure. This includes postgres and the persistent volume it uses.

Migrate

Run make migrate to run the database migration job

Accessing Services

Run make port-forward to forward the ports for all services. The services can then be accessed at the following locations:

• localhost:4000 -> Default service
• localhost:4001 -> Service #1
• localhost:4002 -> Service #2
• localhost:4003 -> Service #3

To cleanup the forwarded ports, run make kill-port-forward. Please note that this will kill all port forwarding, not just those established by make port-forward.

Key code snippets

There are three files involved with setting the environment variables differently for each deployment.

A case is added to server that accepts an argument for which "service" to start. It then sets the SERVICE environment variable before starting the application.

case "$1" in
    "service-1")
        PHX_SERVER=true SERVICE=service-1 exec ./ha_elixir start
        ;;

    "service-2")
        PHX_SERVER=true SERVICE=service-2 exec ./ha_elixir start
        ;;

    "service-3")
        PHX_SERVER=true SERVICE=service-3 exec ./ha_elixir start
        ;;

    *)
        PHX_SERVER=true SERVICE=default exec ./ha_elixir start
        ;;
esac

Inside the application, we simply read from the environment variable and pass it on the the view to be rendered.

def index(conn, _params) do
    render(conn, "index.html", service: System.get_env("SERVICE", "Default"))
end

Finally, we tell kubernetes in k8s/deployments.yaml to override the default command and arguments in order to start the service of our choice.

spec:
    containers:
       - image: ha_elixir
         name: service-1 
         command: ["/app/bin/server"]
         args: ["service-1"]