Treadgold/pyk8sdemo-app

Demo app in Python (Flask) for deployment in Kubernetes

• Flask Demo App
  • Building
  • Pushing the build to a registry
  • Manual Deployment in Kubernetes
    • Preparing a dedicated test namespace (optional, not required when using kustomization)
    • Pod Only Deployment
      • Not using kustomization
      • Using kustomization
      • Checking the deployment
    • Service With Ingress Deployment
      • Not using kustomization
      • Using kustomization
    • Obtaining a Shell from one of the Pods
    • Cleanup
  • Deployment with ArgoCD

Flask Demo App

This is a simple Flask demo app intended to be deployed in Kubernetes

I used this demo app in the following blog articles:

• 2022-04-03
• 2022-04-10

Note: If any of the links above does not work, it may because thearticle is not yet published. You can also visit the blog repository to see work in progress.

Building

Run:

python3 -m build

This will create the Python package.

Next, build the docker image:

docker build --no-cache -t demo-flask-app .

Pushing the build to a registry

Create the following environment variables:

• REGISTRY_URL - contains the ECR or similar Docker container registry URL
• VERSION_TAG - The version number, for example 4.0
• APP_TAG - The application tag, for example demo-flask-app
• LOCAL_IMAGE_ID - The ID of the local container image to use as reference

For a Docker registry running locally on IP address 192.168.2.6 you could do the following:

export REGISTRY_URL=192.168.2.6:5000
export APP_TAG="demo-flask-app"
export LOCAL_IMAGE_ID=`docker image ls | grep demo-flask-app | awk '{print $3}'`
export VERSION_TAG="0.0.1"

To use docker hub, for example, change the REGISTRY_URL and APP_TAG accordingly.

To get the latest image tag info:

docker image list | grep demo-flask-app

Update the environment variable LOCAL_IMAGE_ID

If not using a local image repository, ensure you are logged in. For ECR, for example, run the following:

aws ecr get-login-password | docker login --username AWS --password-stdin $REGISTRY_URL

Then tag and push:

docker tag $LOCAL_IMAGE_ID $REGISTRY_URL/$APP_TAG\:$VERSION_TAG
docker push $REGISTRY_URL/$APP_TAG\:$VERSION_TAG

Manual Deployment in Kubernetes

The manifests are all located in the kubernetes_manifests/ directory of this project and commands will be run from within this directory.

Preparing a dedicated test namespace (optional, not required when using kustomization)

It is not required, but in some cases it may be useful to conduct tests in a temporary namespace.

Follow these steps to create a new namespace and setting it as your default for the kubectl command:

To list the current namespaces, run kubectl get namespaces. Ensure the namespace you intend to use is not already listed.

Now create a namespace with the command kubectl create namespace test

And set it as your default namespace with the command:

kubectl config set-context --current --namespace=test

To verify the current namespace, run the command:

kubectl config view -o jsonpath='{.contexts[].context.namespace}'

The output should be the same namespace name. Any further kubectl commands will be issued against this namespace, unless it is overridden.

The command kubectl get all should also return the following output:

No resources found in test namespace.

Note: The kustomization manifest will create a namespace called demo

Pod Only Deployment

This deployment is useful for scenarios where an ingress is not required. Typical use cases include:

• Troubleshooting or experimenting with ConfigMaps, Secrets, Environment Variables and other internal configurations in the context of a namespace
• Outbound networking tests and troubleshooting (basic tools for these have been included in the image)

Not using kustomization

To deploy the pod only manifest, run the following command:

kubectl apply -f troubleshooting-app-pod-only.yaml

Using kustomization

To deploy the pod only manifest, run the following command:

cd kubernetes_manifests/kustomization-demo

kustomize build overlays/pod-only | kubectl apply -f -

Checking the deployment

For more info, try running kubectl get all -o wide or for maximum information, try running kubectl get all -o yaml

Service With Ingress Deployment

Not using kustomization

The steps are the same as for the Pod Only Deployment with the only difference being to use the troubleshooting-app.yaml manifest file.

Note: It may take 15 to 20 minutes for the Load Balancer to be fully available and the DNS name set. Once this is done, the simple web app can also be opened by a Web Browser, using the DNS name specified in the manifest.

Using kustomization

Run the following command:

cd kubernetes_manifests/kustomization-demo

kustomize build overlays/full | kubectl apply -f -

Obtaining a Shell from one of the Pods

To verify the pod name, run the following command: kubectl get pods

Copy the NAME and run the following command (replacing the pod-name with the one copied):

kubectl exec --stdin --tty pod-name -- /bin/bash

Once in a shell within the pod, troubleshooting activities can be executed.

Cleanup

To delete the deployment, simply run kubectl delete -f troubleshooting-app-pod-only.yaml (or whatever manifest file was used to create the deployment)

To verify, wait a minute or so and run kubectl get all. The output should be something similar to No resources found in test namespace.

Deployment with ArgoCD

The ArgoCD Application Manifest is in argocd/application-manifests/demo1_application.yaml