Operationalize a Machine Learning Microservice API

Project 4 [Udacity Cloud DevOps Engineer Nanodegree]

Project Overview

Deploy a containerized Python flask application to serve out predictions (inference) about housing prices through API calls. It uses a a pre-trained, sklearn model that has been trained to predict housing prices in Boston according to several features.

---

Install

Setup the Environment

• Create a virtualenv and activate it

python3 -m venv <your_venv>
source <your_venv>/bin/activate

• Run make install to install the necessary dependencies

Running app.py

1. Standalone: python app.py
2. Run in Docker: ./run_docker.sh
3. Run in Kubernetes: ./run_kubernetes.sh

Kubernetes Steps

• Setup and Configure Docker locally
• Setup and Configure Kubernetes locally
• Create Flask app in Container
• Run via kubectl

Run the project

1. You should have a virtual machine like VirtualBox and minikube installed, as per the project environmet instructions. To start a local cluster, type the terminal command:

minikube start

2. To deploy this application in kubernetes run:

./run_kubernetes.sh

3. After you’ve called run_kubernetes.sh, and a pod is up and running, make a prediction using a separate terminal tab and run

./make_prediction2.sh

4. Delete the cluster

minikube delete

Files explanation

• config.yml: CircleCI configuration file for running the tests
• app.py: Python flask app that serves out predictions (inference) about housing prices through API calls
• Dockerfile: Dockerfile for building the image
• make_prediction.sh: Send a request to the Python flask app to get a prediction, for localhost
• make_prediction2.sh: Send a request to the Python flask app to get a prediction, for minikube kubernetes
• Makefile: includes instructions on environment setup and lint tests
• run_docker.sh: file to be able to get Docker running, locally
• run_kubernetes.sh: file to run the app in kubernetes
• upload_docker.sh: file to upload the image to docker