KEDA Autoscaling with Nginx

A demonstration of Kubernetes Event-Driven Autoscaling (KEDA) with Nginx deployment, featuring automated load testing and real-time scaling monitoring.

📋 Overview

This project demonstrates horizontal pod autoscaling using KEDA based on CPU utilization. The setup automatically scales an Nginx deployment between 3-10 replicas when CPU usage exceeds 40%.

🚀 Features

  • CPU-Based Autoscaling: Scales based on 40% CPU utilization threshold
  • Configurable Replica Range: 3 minimum, 10 maximum pods
  • Intelligent Cooldown: 30-second cooldown period between scaling events
  • Load Testing Script: Python-based concurrent load generator
  • Real-Time Monitoring: Live pod count and request distribution tracking
  • Even Load Distribution: Kubernetes service ensures balanced traffic across pods

📁 Configuration

KEDA ScaledObject (nginx-scaler.yaml)

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: nginx-scaler
spec:
  scaleTargetRef:
    name: nginx
  minReplicaCount: 3
  maxReplicaCount: 10
  cooldownPeriod: 30
  pollingInterval: 10
  triggers:
    - type: cpu
      metricType: Utilization
      metadata:
        value: "40"

Key Parameters

Parameter Value Description
minReplicaCount 3 Minimum number of pods
maxReplicaCount 10 Maximum number of pods
cooldownPeriod 30s Wait time before scaling down
pollingInterval 10s Frequency of metric checks
cpu.value 40% CPU threshold for scaling

📦 Baseline Setup

Initial Deployment State

Before load testing, the environment was configured with:

Pods (3 replicas at minimum):

kubectl get pods
NAME                    READY   STATUS    RESTARTS   AGE
nginx-8974df4cf-ljcjn   1/1     Running   0          2d1h
nginx-8974df4cf-xm6dz   1/1     Running   0          2d1h
nginx-8974df4cf-ztlsm   1/1     Running   0          2d1h

Service (LoadBalancer):

kubectl get svc
NAME    TYPE           CLUSTER-IP     EXTERNAL-IP   PORT(S)        AGE
nginx   LoadBalancer   10.43.212.55   10.10.10.50   80:30858/TCP   2d2h

The LoadBalancer exposes Nginx at 10.10.10.50:80, which serves as the target for load testing.

🧪 Load Testing

Running the Test

uv run nginx-keda.py http://10.10.10.50 240 200

Arguments:

  • http://10.10.10.50 - Target URL
  • 240 - Duration in seconds
  • 200 - Concurrent users

Real-Time Scaling Events

The load test script monitors pod scaling in real-time:

======================================================================
KEDA AUTOSCALING LOAD TEST
======================================================================
Target URL: http://10.10.10.50
Duration: 240 seconds
Concurrent Users: 200
Initial Pod Count: 3
======================================================================
======================================================================
MONITORING POD SCALING
======================================================================
Time (s)   Pods       Requests     Errors     HPA Status
----------------------------------------------------------------------
35         6          17923        0          3     6
🚀 SCALING EVENT: 3 -> 6 pods
User  3 completed: 543 requests    128        6     6
User  2 completed: 560 requests

Scaling Behavior

The test demonstrates effective autoscaling:

  • Initial State: 3 pods
  • Scaling Trigger: CPU utilization exceeded 40% threshold
  • Scale Up Time: 35 seconds (3 → 6 pods)
  • Requests at Scale Event: 17,923 requests processed
  • Final State: 6 pods maintained throughout test
  • Load Distribution: Even distribution across all pods (~15k-21k requests per pod)

Test Results

======================================================================
LOAD TEST RESULTS
======================================================================
Test Duration: 251.9 seconds
Total Requests: 107830
Successful: 107702
Errors: 128
Requests/sec: 428.07
Final Pod Count: 6
======================================================================

📊 Pod Distribution

Traffic was evenly distributed across scaled pods:

nginx-8974df4cf-ztlsm: 20997 (19.5%) █████████
nginx-8974df4cf-ljcjn: 20854 (19.4%) █████████
nginx-8974df4cf-xm6dz: 20832 (19.3%) █████████
nginx-8974df4cf-h8pbd: 15073 (14.0%) ██████
nginx-8974df4cf-248n5: 15061 (14.0%) ██████
nginx-8974df4cf-t6htr: 14885 (13.8%) ██████

🔍 Monitoring Commands

Check HPA Status

kubectl get hpa keda-hpa-nginx-scaler

Output:

NAME                    REFERENCE          TARGETS       MINPODS   MAXPODS   REPLICAS   AGE
keda-hpa-nginx-scaler   Deployment/nginx   cpu: 0%/40%   3         10        3          2d1h

Watch Pods Scale

kubectl get pods -l app=nginx -w

View Current Pods

kubectl get pods

🛠️ Setup

Prerequisites

  • Kubernetes cluster
  • KEDA installed (installation guide)
  • kubectl configured
  • Python 3.x with uv (for load testing)

Installation

  1. Deploy the ScaledObject:

    kubectl apply -f nginx-scaler.yaml

  2. Verify KEDA Configuration:

    kubectl get scaledobject nginx-scaler

  3. Check HPA Creation:

    kubectl get hpa

💡 Key Insights

  • Responsive Scaling: Pods scaled up in 35 seconds under load
  • Load Balancing: Kubernetes service distributed ~108k requests evenly
  • Stability: 99.88% success rate (107,702/107,830 requests)
  • Performance: Sustained 428 requests/second with 200 concurrent users

📚 Additional Resources