Kubernetes for ML workloads in MLOps - Mini Project: Build & Apply

Kubernetes for ML Workloads

📖 Scenario: You are a data scientist who wants to run a machine learning training job on Kubernetes. You will create a simple Kubernetes Pod configuration to run a Python script that trains a model. This project will guide you step-by-step to create the YAML configuration, add resource limits, and finally deploy and check the Pod status.

🎯 Goal: Build a Kubernetes Pod YAML file to run a machine learning training script, add resource limits, and deploy it to see the Pod running.

📋 What You'll Learn

Create a basic Kubernetes Pod YAML file named ml-training-pod.yaml with a container running Python

Add resource limits for CPU and memory to the container

Deploy the Pod using kubectl apply

Check the Pod status using kubectl get pods

💡 Why This Matters

🌍 Real World

Data scientists and ML engineers use Kubernetes to run training jobs reliably and scale them easily in production environments.

💼 Career

Knowing how to configure and deploy ML workloads on Kubernetes is a key skill for MLOps engineers and DevOps professionals working with AI projects.

1

Create the basic Pod YAML

Create a file named ml-training-pod.yaml with a Kubernetes Pod configuration. The Pod should be named ml-training-pod and run a container named ml-container using the image python:3.12-slim. The container should run the command python with arguments -c and print('Training started').

MLOps

# Create ml-training-pod.yaml with the Pod spec
# Your code here

Need a hint?

Remember to use YAML indentation carefully. The container spec goes under spec.containers.

2

Add resource limits to the container

In the existing ml-training-pod.yaml file, add resource limits to the container ml-container. Set the CPU limit to 500m and memory limit to 256Mi under resources.limits.

MLOps

apiVersion: v1
kind: Pod
metadata:
  name: ml-training-pod
spec:
  containers:
  - name: ml-container
    image: python:3.12-slim
    command: ["python"]
    args: ["-c", "print('Training started')"]
    # Add resource limits below
    # Your code here

Need a hint?

Resource limits go under the container spec with indentation. Use quotes around values like "500m".

3

Deploy the Pod to Kubernetes

Use the command kubectl apply -f ml-training-pod.yaml to deploy the Pod to your Kubernetes cluster.

MLOps

# Run the kubectl apply command below
# Your code here

Need a hint?

This command tells Kubernetes to create or update resources defined in the YAML file.

4

Check the Pod status

Use the command kubectl get pods to check the status of the Pod named ml-training-pod. The output should show the Pod with status Running or Completed.

MLOps

# Run the kubectl get pods command below
# Your code here

Need a hint?

Look for the Pod name ml-training-pod in the list and check its STATUS column.