DevOps Monitoring Hands-On: Prometheus & Grafana on Kubernetes Kind Cluster Using Helm by Shiva Sir

This is simplest tutorial for DEVOPS Monitoring Tools:

follow this repository of github for source code and project

github repository for kind_cluster , graphna

 Modern applications like Netflix, Hot star, and large-scale SaaS platforms survive because of real-time monitoring and observability. Without continuous visibility into traffic, CPU usage, memory, and failures, production systems become blind.

In this hands-on guide, we implement end-to-end Kubernetes monitoring using:

• Prometheus – metrics scraping & storage

• Grafana – visualization & dashboards

• Kubernetes – orchestration platform

• Helm – package management

• Amazon EC2 – infrastructure

• Kind (Kubernetes in Docker) – local cluster simulation

This setup mirrors real production practices while remaining cost-effective and beginner-friendly.

---

Why Monitoring & Observability Matter

Monitoring answers:

• What is happening now?

Observability answers:

• Why is it happening?

Using Prometheus + Grafana, we monitor:

• Node health

• Pod & container metrics

• CPU, memory, disk, network

• Application load

• Kubernetes control-plane components

---

Architecture Overview

High-level flow:

1. Node Exporter (DaemonSet) runs on every node

2. kube-state-metrics collects cluster state

3. Prometheus scrapes metrics

4. Grafana visualizes metrics via dashboards

5. Helm installs everything as a unified stack

---

Infrastructure Setup

Step 1: Launch EC2 Instance

• Instance type: t2.large

• OS: Ubuntu

• Storage: Minimum 25 GB

• Open inbound ports:

  • 9090 – Prometheus

  • 3000 – Grafana

---

Prerequisites Installation

We install everything using a shell script:

Required Tools

• Docker

• kubectl

• Kind

• Helm

chmod +x install.sh
./install.sh

Verify:

docker --version
kubectl version --client
kind version
helm version

---

Creating Kubernetes Cluster Using Kind

Kind Configuration

• 1 Control Plane

• 3 Worker Nodes

• Port mappings enabled

kind create cluster \
  --name aim-cluster \
  --config kind-config.yaml

Check:

kubectl get nodes
docker ps

---

Installing Prometheus & Grafana Using Helm

Step 1: Create Monitoring Namespace

kubectl create namespace monitor

Step 2: Add Prometheus Helm Repository

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
helm repo update

Step 3: Install Monitoring Stack

helm install prometheus-stack \
  prometheus-community/kube-prometheus-stack \
  --namespace monitor \
  --set grafana.service.type=NodePort \
  --set prometheus.service.type=NodePort

---

What Helm Deploys Automatically

Component	Purpose
Prometheus	Metrics storage
Grafana	Visualization
Alertmanager	Alert handling
Node Exporter	Node metrics
kube-state-metrics	Cluster state
DaemonSets	One exporter per node
StatefulSets	Alertmanager

---

Accessing Grafana & Prometheus

Retrieve Grafana Password

kubectl get secret \
  --namespace monitor \
  prometheus-stack-grafana \
  -o jsonpath="{.data.admin-password}" | base64 --decode

• Username: admin

• Password: (decoded value)

Port Forwarding

kubectl port-forward svc/prometheus-stack-grafana 3000:80 -n monitor
kubectl port-forward svc/prometheus-stack-kube-prometheus-prometheus 9090:9090 -n monitor

Access in browser:

• Grafana → http://EC2-IP:3000

• Prometheus → http://EC2-IP:9090

---

Grafana Dashboards

Grafana automatically includes:

• Kubernetes cluster dashboards

• Node performance

• Pod resource usage

• Network traffic

• CPU & memory trends

You can also import dashboards using Grafana Dashboard IDs (e.g., 15661).

---

Deploying Sample Voting Application

To demonstrate real-time application monitoring, a microservices-based Voting App is deployed:

Stack Used

• Frontend: Python / Node.js

• Cache: Redis

• Database: PostgreSQL

kubectl apply -f k8s-specifications/

Services:

• Vote (NodePort)

• Result (NodePort)

• Redis & DB (ClusterIP)

---

Observing Real-Time Load

As users vote:

• CPU spikes visible

• Memory usage increases

• Network traffic grows

• Node-level metrics change in real time

This is exactly how production monitoring works.

---

Key Kubernetes Concepts Demonstrated

Concept	Usage
DaemonSet	Node Exporter on every node
Deployment	Application workloads
StatefulSet	Alertmanager
NodePort	External access
ClusterIP	Internal services
Namespace	Isolation (monitor)

---

Why Helm Is Critical in Real Projects

Without Helm:

• Hundreds of YAML files

• Manual updates

• High error risk

With Helm:

• One command deployment

• Version control

• Rollbacks

• Production-grade standardization

---

Cleanup & Cost Control

helm uninstall prometheus-stack -n monitor
kind delete cluster --name aim-cluster

Important: Terminate EC2 instance to avoid AWS charges.