The repository showcases how to operationalize a machine learning microservice API with Docker and deploy it at scale using Kubernetes. The application consists of a pre-trained machine learning model built on sklearn and geared towards predicting house prices in the city of Boston. The focus of this project is on said operationalization and deployment. Therefore, the actual application being deployed has been provided by a third party.
Before everything can be run on your local machine a few things need to be in place. Please make sure that Docker is installed and running locally. Download it for your system here. Additionally you need to have minikube as well as kubectl configured in order to set up and run a local Kubernetes cluster. Learn more about minikube and kubectl.
A simple Makefile has been created to get you up and running quickly. Find a comprehensive list of the commands available through said Makefile below.
| Command | Description |
|---|---|
| make setup | Create proper python venv |
| make install | Install all necessary dependencies |
| make validate-circleci | Validate CircleCI configuration file |
| make run-circleci | Run CircleCI locally to verify build |
| make lint | Lint Docker container and Python scripts |
A variety of bash scripts are provided to streamline certain tasks related to Docker containers and Kubernetes clusters.
| Bash Script | Description |
|---|---|
| run_docker.sh | Builds and runs Docker container with port 80 exposed and forwarded to host port 8000 |
| upload_docker.sh | Authenticates with Docker Hub and uploads finished Docker container |
| run_kubernetes.sh | Runs Docker container on local Kubernetes cluster with port 80 forwarded to host port 8000 |
| make_prediction.sh | Runs test prediction to validate if everything is up and running correctly |
There are few other important files to consider when working with this repo. The table below gives a short overview.
| File | Description |
|---|---|
| app.py | Contains the actual machine learning model to be operationalized |
| requirements.txt | List of dependencies that need to be installed in container for the app to function properly |
| Dockerfile | Configures how the Docker container should be built |
The Docker Hub repository with the container created and deployed here can be found here.
