Predictionapp is a project centred around operationalizing a machine learning microservice using Kubernetes, an open-source system for automating deployment, scaling, and management of containerized applications. In this project, a pre-trained model designed to predict Boston housing prices gotten from the Boston Kaggle dataset is used. The pre-trained model is wrapped in a Flask app and predictions are made from API calls made to the app. The Flask app is further containerized using Docker and Kubernetes. Automated tests are made using CircleCI.

1) Open your terminal and clone the repo using the following:

git clone https://github.com/Amamgbu/predictionapp.git

2) Navigate into directory:

cd predictionapp

3) Make sure python is already installed in your PC. If not already installed, download latest version here: python

4) Create a virtual environment and activate by running the following command:

make setup

5) In order to install project dependencies, run the following:

make install

6) Lint files to test for syntax correctness using command:

make lint

7) In order to build docker image and start app, run the following command:

./run_docker.sh

8) In order to test app by making predictions, run the command:

./make_prediction.sh

Once predictions are made successfully, the following processes are followed:

9) Edit upload_docker.sh and change dockerpath to contain ID(username) of your docker container. After this is done, run the following:

./upload_docker.sh

10) The following command is run in order to run the service in a Kubernetes cluster. Ensure Minikube and Kubernetes is installed on your local machine. Follow this tutorial for more info on how to achieve this on Ubuntu: [Kubernetes Tutorial:Ubuntu] (https://phoenixnap.com/kb/install-minikube-on-ubuntu). After install, edit dockerpath on run_kubernetes.sh and replace with your DockerID.

./run_kubernetes.sh

11) You can test the app deployed by making predictions. Make sure port forwarding is successful.

./make_prediction.sh

• app.py: This contains the python code that runs the Flask app.
• Makefile: This contains code with commands for setup, installation and lint of project.
• Dockerfile: This file sets up Docker image.
• run_docker.sh: Shell script responsible for creating a docker image from the Dockerfile, listing the images and starting a docker container
• upload_docker.sh: This script uploads a docker image to docker hub of ID provided.
• run_kubernetes.sh: Shell script responsible for running image in Kubernetes cluster.
• make_prediction.sh: Shell script responsible for testing deployed microservice with dummy data in order to make predictions.