Kubernetes Learning Notes & Practical Workflow

A comprehensive guide covering Kubernetes fundamentals, CI/CD integration, managed Kubernetes services (EKS, AKS, GKE), Minikube setup, deployment workflow, troubleshooting, and production best practices.

---

Table of Contents

• Introduction
• Kubernetes in CI/CD Workflow
• Where Kubernetes Clusters are Created in Real Life
• Managed Kubernetes Services
  • AWS EKS
  • Azure AKS
  • Google GKE
• EKS vs AKS vs GKE Comparison
• Kubernetes Project Workflow
• Minikube Installation
• Deployment Steps
• Troubleshooting
• Debugging
• Load Testing
• ImagePullBackOff Fix
• Key Takeaways

---

Introduction

Kubernetes is an open-source container orchestration platform that automates:

• Deployment
• Scaling
• Load balancing
• Self-healing
• Rolling updates

Instead of manually running Docker containers on servers, Kubernetes manages containers automatically.

---

Kubernetes in a CI/CD Workflow

Traditional Workflow

GitHub
    ↓
GitHub Actions
    ↓
Build Docker Image
    ↓
Push to Amazon ECR
    ↓
Deploy to EC2

---

Kubernetes Workflow

GitHub
    ↓
GitHub Actions
    ↓
Build Docker Image
    ↓
Push Image to ECR
    ↓
kubectl apply
    ↓
Kubernetes Cluster

Instead of deploying directly on an EC2 instance, GitHub Actions deploys the latest Docker image into a Kubernetes cluster.

Example deployment step:

- name: Deploy to Kubernetes
  run: |
    kubectl apply -f deployment.yaml

---

Example GitHub Actions Workflow

name: CI/CD Pipeline with Kubernetes

on:
  push:
    branches:
      - main

jobs:
  deploy:
    runs-on: ubuntu-latest

    steps:

      - uses: actions/checkout@v3

      - name: Login to ECR
        run: aws ecr get-login-password --region <region> |
             docker login --username AWS --password-stdin \
             <account-id>.dkr.ecr.<region>.amazonaws.com

      - name: Build Docker Image
        run: |
          docker build -t app:latest .
          docker push <ecr-url>/app:latest

      - name: Deploy to Kubernetes
        uses: azure/k8s-deploy@v4
        with:
          manifests: |
            deployment.yaml
            service.yaml

---

Benefits of Kubernetes

• Automatic scaling
• Rolling updates
• Zero downtime deployment
• Self healing
• Load balancing
• High availability

---

Where Are Kubernetes Clusters Created?

Local Development

Used only for learning and testing.

Tools:

• Minikube
• Kind
• Docker Desktop Kubernetes

---

Production

Companies generally use managed Kubernetes services instead of self-hosting.

Popular options:

• Amazon EKS
• Azure AKS
• Google GKE

Large enterprises with strict compliance requirements may also create on-premise Kubernetes clusters using tools like:

• kubeadm
• Rancher

---

Why Minikube Isn't Used in Production

Minikube is intended only for local development.

Limitations:

• Single-machine setup
• Limited scalability
• No production-grade high availability
• Not suitable for enterprise workloads

---

Managed Kubernetes Services

Cloud providers manage the Kubernetes control plane while developers manage their applications.

---

AWS Elastic Kubernetes Service (EKS)

Features

• Managed Control Plane
• IAM integration
• CloudWatch integration
• Elastic Load Balancer support
• VPC networking
• Auto Scaling Groups

Best For

• Existing AWS users
• Enterprise applications
• Large-scale workloads

---

Azure Kubernetes Service (AKS)

Features

• Beginner-friendly
• Azure AD integration
• Azure Monitor
• Blob Storage support
• Automated upgrades
• Virtual Nodes

Best For

• Microsoft ecosystem
• Azure ML
• Hybrid cloud deployments

---

Google Kubernetes Engine (GKE)

Features

• Created by Google
• GKE Autopilot
• Automatic scaling
• Cloud Monitoring
• Vertex AI integration
• GPU and TPU support

Best For

• AI/ML workloads
• Startups
• Modern cloud-native applications

---

EKS vs AKS vs GKE

Feature	AWS EKS	Azure AKS	Google GKE
Ease of Use	Medium	Beginner Friendly	Beginner Friendly
Cloud Integration	AWS	Azure	Google Cloud
AI/ML Support	SageMaker	Azure ML	Vertex AI
Networking	Advanced	Simple	Simple
Autoscaling	Auto Scaling Groups	Built-in	Advanced
Best For	Enterprise	Microsoft Users	AI/ML

---

Kubernetes Project Workflow

---

Step 1

Build your application.

Verify it works locally.

---

Step 2

Create Docker Image

docker build -t kubernetes-test-app:latest .

Verify:

docker images

Run locally:

docker run -p 5000:5000 kubernetes-test-app:latest

---

Step 3

Create

deployment.yaml

---

Step 4

Install Minikube

Start Minikube

minikube start

or

minikube start --embed-certs

---

Verify Minikube

minikube status

kubectl get all -A

kubectl get pods -A

kubectl get nodes -A

---

Add Multiple Nodes

minikube start --nodes=2

or

minikube start --nodes=2 --embed-certs

---

Load Docker Image into Minikube

List Docker images

docker images

List Minikube images

minikube image list

Load image

minikube image load kubernetes-test-app:latest

---

Deploy Application

kubectl apply -f deployment.yaml

Delete deployment

kubectl delete deployment kubernetes-test-app

---

Test Deployment

Check Pods

kubectl get pods

Check Nodes

kubectl get nodes

Delete Pod

kubectl delete pod <pod-name>

Access Service

minikube service kubernetes-test-app

Open Dashboard

minikube dashboard

---

Debugging

View logs

kubectl logs -f <pod-name>

View endpoints

kubectl get endpoints

View services

kubectl get service

---

Load Testing

Use Postman Performance Runner.

Example:

• 10 users
• 1 minute
• Fixed configuration

Monitor:

• Response time
• Error rate
• Throughput

---

Stop Minikube

minikube stop

---

ImagePullBackOff Fix

Tag image

docker tag kubernetes-test-app:latest \
<dockerhub-username>/kubernetes-test-app:latest

Push image

docker push <dockerhub-username>/kubernetes-test-app:latest

---

Troubleshooting

Registry Connection Error

Failed to connect to registry.k8s.io

If using a proxy:

minikube start \
--docker-env HTTP_PROXY=http://proxy:port \
--docker-env HTTPS_PROXY=https://proxy:port

Without a proxy:

unset HTTP_PROXY
unset HTTPS_PROXY
unset NO_PROXY

Check DNS

nslookup registry.k8s.io

Reset Minikube

minikube stop

minikube delete --all

Restart

minikube start

---

Key Takeaways

• Kubernetes automates deployment and scaling of containerized applications.
• GitHub Actions can deploy directly to Kubernetes clusters using kubectl.
• Minikube is ideal for learning but not suitable for production.
• Managed Kubernetes services (EKS, AKS, and GKE) reduce operational complexity.
• GKE is particularly strong for AI/ML workloads due to GPU/TPU and Vertex AI integration.
• Kubernetes provides rolling updates, self-healing, load balancing, and high availability.
• Production environments typically rely on managed Kubernetes services rather than self-hosted clusters.

---

Learning Roadmap

Docker
    ↓
Docker Compose
    ↓
Kubernetes Basics
    ↓
Pods
    ↓
ReplicaSets
    ↓
Deployments
    ↓
Services
    ↓
ConfigMaps & Secrets
    ↓
Ingress
    ↓
Volumes
    ↓
Helm
    ↓
Kubernetes on Cloud (EKS / AKS / GKE)
    ↓
CI/CD with Kubernetes

---

Author

Ayush

These notes were created as part of an AI/ML and MLOps learning journey, covering Kubernetes from local development with Minikube to production-ready deployments using managed Kubernetes services.