Skip to content

Latest commit

 

History

History
87 lines (61 loc) · 5.09 KB

kube_env_vars.md

File metadata and controls

87 lines (61 loc) · 5.09 KB
Raw

~~ Environment Variables ~~

Redeploy the baseline

One of your original asks was that this application be configurable via Environment Variables. We're now going to put that to the test. Our first steps will be to redeploy the application, now using the declarative YAML in the k8s/app-envvars directory of this project. To do that, we're going to use the kubectl apply ... command.

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

If you check your dashboard, you should see a single Pod running again. Go ahead and take a look at the k8s/app-envvars/deployment-base.yaml file to get familiar with it. It's a slimmed down version of the Deployment YAML you got in the previous step with kubectl get ... -oyaml.

Connect to the Container

We never actually tested the application in the last step, but lets do that now by exposing the Pod as a Service and using the kubectl port-forward command to connect to it. A Kubernetes Service is a lengthy topic, and more can be read here. We'll be focusing on creating what's known as a ClusterIP Service, which means there will be a single addressable endpoint available in the cluster to access the Pod(s) hosting our application. First thing is first, let's expose the Service using kubectl expose ....

kubectl expose deployment cn-demo --port 5000 --target-port 5000

This command says "Create a new ClusterIP Service that listens on port 5000 and targets port 5000 of all Pods that are a part of the cn-demo Deployment". Next lets connect to this newly created Service. The kubectl port-forward command will expose a port on our local machine that connects to ports exposed on Pods. We need to do the port-forward command here because the Pod is only accessible inside the cluster right now.

kubectl port-forward svc/cn-demo 5000

Earlier we said that port-forward connects to a Pod, but here we're port forwarding to a Service. What??? In this case, think of the Service as an alias to one to many Pods that it load balances across. It's a kubectl trick, but just a single Pod will be selected for forwarding, even if there are many replicas of it.

At this point we should be able to access http://localhost:5000 from our browser. Give it a try. You should see the familiar JSON output we saw earlier when running the container with Docker. Your value for host will vary.

{
  "appName": "cloud-native-demo",
  "env": {
    "host": "cn-demo-7df548dcf8-wpr8v",
    "user_defined_1": null,
    "user_defined_2": null,
    "user_defined_3": null
  },
  "redis-host": "localhost",
  "version": "1.0.0"
}

Once you're finished, stop the port forwarding command by pressing CTRL+C.

Configure the User Defined variables

TIP: The completed deployment configuration is in the k8s/app-envvars/deployment-with-envvars.yaml file if you need help through this section.

Right now we have a lot of null values for the user_defined_# variables. Let's fix that. In the k8s/app-envvars/deployment-base.yaml file, find the section that looks like the following:

      containers:
      - image: cloud-native-demo:1
        imagePullPolicy: IfNotPresent
        name: cn-demo

Add a new section directly under name: cn-demo with the following. Indentation matters here.

        env:
          - name: USER_DEFINED_1
            value: my-user-define-value-1

With just this configuration, re-apply this Deployment. Kubernetes will see the difference in configuration and redeploy a new Pod with the updated configuration. The old Pod will be terminated as it no longer matches the desired state of your deployment (it doesn't have the right configuration anymore). Assuming you didn't change the filename, run the following and watch your dashboard for how the old and new Pods cycle.

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

Finally, run the same port-forward command you ran earlier and try to access the application in your browser at http://localhost:5000. Do you see the updated value for user_defined_1? If you look at the application code in the app.py file at the root of the repo you cloned, you can see how the Python code is reading from an Environment Variable to populate that section of the JSON response.

NOTE: Be very careful in a production system with your environment configuration. It's useful for this demo to return environment variables off one of the API endpoints. A production system though will sometimes hold sensitive information in environment variables. Would you want someone seeing your database connection string after calling a URL like this? Nope!

Fill in the remaining environment variables

Using the same pattern as above, fill in the following environment key-value pairs as environment variables and observe the changes in the running application after redeploying.

USER_DEFINED_2=my-user-define-value-2
USER_DEFINED_3=my-user-define-value-3
REDIS_HOST=redis-test
REDIS_PORT="6379"