Machine-Learning-on-Kubernetes

Step 1: Set up a Functional Kubernetes Cluster

1. Open GKE terminal.

2. Start minikube in Google Cloud Platform

     minikube start
   

3. Create requirements.txt file using the following command:

    nano requirements.txt
   

Enter the following contents in requirements.txt:

    Flask==1.1.1
    gunicorn==19.9.0
    itsdangerous==1.1.0
    Jinja2==2.10.1
    MarkupSafe==1.1.1
    Werkzeug==0.15.5
    numpy==1.19.5
    scipy>=0.15.1
    scikit-learn==0.24.2
    matplotlib>=1.4.3
    pandas>=0.19
    flasgger==0.9.4

Upload logreg.pkl file which we got from ML.ipynb and using data set

• Click the three dots in the top-right part of the Cloud Shell Terminal

• Choose upload and upload the logreg.pkl file

4. Create flask_api.py file using the command:

       nano flask_api.py
   

or, you can also uplaod it from flask_api.py file

5. Create Dockerfile using the command:

       nano Dockerfile
   

Enter the following contents in Dockerfile:

    FROM python:3.8-slim
    WORKDIR /app
    COPY . /app
    EXPOSE 5000
    RUN pip install -r requirements.txt
    CMD ["python", "flask_api.py"]

6. Build the Docker image using the command:

       sudo docker build -t ml_app_docker .
   

7. Run the Docker container using the command:

       docker container run -p 5000:5000 ml_app_docker
   

8. Preview the application:

• In the upper-right side of the terminal, click the eye-shaped button and then click "Preview on port 5000".

  

• Add /apidocs/ at the end of the link to access the running ml-app.

10. Make predictions for a group of customers (test data) via a POST request:

• Upload the test data file containing the same parameters in a similar order.

• Execute to display the results.

  Step 1.1. Stop/kill the running container:

• List running Docker containers using the command:

     docker ps
  

• Use the command to kill the running container

    docker kill <CONTAINER ID>
  

  Step 2: Push the image to your docker hub

1. Log in to Docker Hub:

      docker login
   

2. Tag the Image:

     docker tag ml_app_docker nazninnahar054/ml_docker:latest
   

3. Push the Image:

     docker push nazninnahar054/ml_docker:latest
   

4. Verify the image exists locally by listing your Docker images:

     docker images
   

P.S. Before login to docker build the image if pushing isn't working

Step 3: Deploy your ML app to GKE

Use the GKE we have created in Step 1

1. Create a deployment.yaml with the following contents.

     nano deployment.yaml
   

   then write this content

2. Create deployment with the above file.

     kubectl apply -f ml-app-deployment.yaml
   

3. Wait for couple minutes and list all the pods created

     kubectl get deployments
     kubectl get pods
   

4. Create a service.yaml

     nano service.yaml
   

then write this content

6. Create service with the above file

     kubectl apply -f ml-app-service.yaml
   

7. Get service external ip

     kubectl get services
   

8. Access using browser:

     external-ip/apidocs