A simple walkthrough and template for kubernetes (k8s) flask app deployments on the AWS cloud. The solution leverages AWS' Elastic K8s Service (EKS) to manage clusters on the public cloud. The code here provides both the infrastructure as code (IaC) components and a simple application script to handle the flask API.
This folder contains the code for the application image.
- Dockerfile contains the instructions for building an image
- app.py contains the flask app
- requirements.txt has all of the dependancies for the application
This folder contains the kubernetes and EKS configurations.
- cluster.yml EKS cluster configuration script
- models-deployment.yml K8s deployment script
- models-service.yml K8s service script
The diagram below outlines the major components of the solution.
- Use docker cli to build image locally
docker build -t <image_tag> . - Install AWS cli and ensure you set permissions with
aws configure - Create ECR registry and push image to ECR repo
- Install eksctl and kubectl
- Create cluster config file
- Configure and launch cluster with eksctl
eksctl create cluster -f cluster.yaml - Create kubernetes namespace
kubectl create namespace <namespace> - Create kubernetes deployment and service .yml files
- Create deployment using
kubectl apply -f <filename> - Create service using
kubectl apply -f <filename> - Check for status of deployment and service using
kubectl -n <namespace> get all - Retrieve load balancer external IP and make a HTTP request to test app
- EKS resources such as pods, replica groups, and nodes won't show up in AWS console unless you a IAM policy that gives you access. EKS resource details are not available on aws cli or eks sdk.
- To force update pods with new image, update the name of the container in the deployment file and apply again.
- If you want to access as a pod shell you can use
kubectl exec -it <pod id> -n <namespace> -- bash(not recommended to have bash or powershell installed on production images)