kube_override_cmd.md

Pass the REDIS_HOST env var as a CLI argument to the application

While our application has the ability to read directly from the configured Environment Variables, some applications are configured via CLI arguments on startup. So far, we haven't actually told Docker or Kubernetes which Python script to run at startup. It's been using the default CMD instruction defined in the original Dockerfile it was built from. You can see in the Dockerfile in this repository, it's running the command python app.py by default.

This program actually accepts a few configuration values on the CLI. Our always helpful development team left a --help argument in there to help discover what some of them are. Let's run that now in a new container using Docker. It only needs to run long enough to give us the output. In a new console window, run the following:

docker run --rm cloud-native-demo:1 python app.py --help

You should see the following output:

docker run --rm cloud-native-demo:1 python app.py --help

usage: app.py [-h] [--port PORT] [--host HOST] [--redis-host REDIS_HOST]
              [--redis-port REDIS_PORT]

Starter Python + Redis app

optional arguments:
  -h, --help            show this help message and exit
  --port PORT, -p PORT  port number to listen for HTTP requests
  --host HOST           host to bind to
  --redis-host REDIS_HOST
                        hostname of the backing Redis service
  --redis-port REDIS_PORT
                        port number of the backing Redis service

Soon, we're going to deploy an instance of Redis into the cluster to connect to. Let's pass the REDIS_HOST environment variable we configured in the last step as a CLI argument to the Python application so we can control the Redis host the application connects to.

You can override the start command (right now it's defaulting to what was defined in the Dockerfile) using the command property in your Deployment YAML. The command property goes at the same depth the env property did in the last step. Add the following to your Deployment and try to redeploy it with the kubectl apply -f ... command used before. Take note of how REDIS_HOST is passed, surrounded by $(). This is important for the variable to resolve at runtime, otherwise it would just take the literal string $REDIS_HOST.

Go ahead and update the k8s/app-envvars/deployment-base.yaml now to override the command the container should run. See the syntax below.

Use the completed deployment-with-envvars.yaml file as a guide if you're having trouble placing this configuration in the file.

        command:
          - sh
          - -c
          - |
            python app.py --redis-host $(REDIS_HOST)

After updating the yaml file, go ahead and test it again. If you have pods already running, you may want to delete your deployment first by using the kubectl delete deploy command:

kubectl delete deploy <deployment name>

Now create the deployment with your new overridden command by using the kubectl apply command:

kubectl apply -f k8s/app-envvars/deployment-base.yaml

Watch your console to see your new pod start up. Once you have a new running pod, we will use the port forward command to connect to the pod.

Last time we created a service to expose your deployment, but if need be you can port-forward directly into a single pod like this:

kubectl port-forward pod/<podname> 5000:5000

Port forward to the pod and then test the application by hitting the http://localhost:5000 endpoint again.

You should see a similar JSON response as last time, except this time you should see the redis host has the value you specified as an environment variable in your deployment definition.