KEDA on Kubernetes: Event-Driven Autoscaling from Zero to Hero

Why KEDA?

The built-in Kubernetes HPA scales on CPU and memory. But real workloads scale on:

• Messages in a queue (RabbitMQ, Kafka, SQS)
• Pending HTTP requests
• Database connection count
• Custom business metrics
• Time of day (cron-based scaling)

KEDA (Kubernetes Event-Driven Autoscaling) adds 60+ scalers and the ability to scale to zero — something native HPA cannot do.

Installation

helm repo add kedacore https://kedacore.github.io/charts
helm repo update

helm install keda kedacore/keda \
  --namespace keda \
  --create-namespace \
  --set webhooks.enabled=true

Verify:

kubectl get pods -n keda
# keda-operator-xxx         Running
# keda-metrics-apiserver-xxx Running
# keda-admission-webhooks-xxx Running

Core Concepts

KEDA introduces two CRDs:

• ScaledObject — scales Deployments/StatefulSets
• ScaledJob — scales Jobs (batch processing)

┌─────────────┐     ┌──────────┐     ┌────────────────┐
│ Event Source │────▶│   KEDA   │────▶│ HPA (under hood)│
│ (Queue/API) │     │ Operator │     │ scales workload │
└─────────────┘     └──────────┘     └────────────────┘

Example 1: Scale on RabbitMQ Queue Depth

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: order-processor
spec:
  scaleTargetRef:
    name: order-processor  # Deployment name
  pollingInterval: 10
  cooldownPeriod: 30
  minReplicaCount: 0      # Scale to zero when queue is empty
  maxReplicaCount: 50
  triggers:
    - type: rabbitmq
      metadata:
        host: amqp://user:pass@rabbitmq.default:5672/
        queueName: orders
        queueLength: "5"   # 1 pod per 5 messages

When the orders queue has 0 messages → 0 pods. 25 messages → 5 pods. 250 messages → 50 pods.

Example 2: Scale on Kafka Consumer Lag

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: event-consumer
spec:
  scaleTargetRef:
    name: event-consumer
  minReplicaCount: 1
  maxReplicaCount: 100
  triggers:
    - type: kafka
      metadata:
        bootstrapServers: kafka-broker:9092
        consumerGroup: my-consumer-group
        topic: events
        lagThreshold: "10"
        offsetResetPolicy: earliest

Example 3: Scale on Prometheus Metrics

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: api-gateway
spec:
  scaleTargetRef:
    name: api-gateway
  minReplicaCount: 2
  maxReplicaCount: 20
  triggers:
    - type: prometheus
      metadata:
        serverAddress: http://prometheus.monitoring:9090
        metricName: http_requests_per_second
        query: |
          sum(rate(http_requests_total{app="api-gateway"}[1m]))
        threshold: "100"   # 1 pod per 100 req/s

Example 4: Cron-Based Scaling

Pre-scale for known traffic patterns:

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: web-frontend
spec:
  scaleTargetRef:
    name: web-frontend
  minReplicaCount: 2
  maxReplicaCount: 30
  triggers:
    - type: cron
      metadata:
        timezone: Europe/Amsterdam
        start: 0 8 * * 1-5    # 8 AM weekdays
        end: 0 20 * * 1-5     # 8 PM weekdays
        desiredReplicas: "10"
    - type: cron
      metadata:
        timezone: Europe/Amsterdam
        start: 0 20 * * 1-5   # 8 PM weekdays
        end: 0 8 * * 1-5      # 8 AM weekdays
        desiredReplicas: "3"

Example 5: AWS SQS

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: sqs-processor
spec:
  scaleTargetRef:
    name: sqs-processor
  minReplicaCount: 0
  maxReplicaCount: 25
  triggers:
    - type: aws-sqs-queue
      metadata:
        queueURL: https://sqs.us-east-1.amazonaws.com/012345678901/my-queue
        queueLength: "5"
        awsRegion: us-east-1
      authenticationRef:
        name: aws-credentials
---
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: aws-credentials
spec:
  secretTargetRef:
    - parameter: awsAccessKeyID
      name: aws-secret
      key: access-key
    - parameter: awsSecretAccessKey
      name: aws-secret
      key: secret-key

ScaledJobs for Batch Processing

Instead of long-running pods, spin up Jobs per message:

apiVersion: keda.sh/v1alpha1
kind: ScaledJob
metadata:
  name: image-processor
spec:
  jobTargetRef:
    template:
      spec:
        containers:
          - name: processor
            image: image-processor:latest
            env:
              - name: QUEUE_URL
                value: amqp://rabbitmq:5672
        restartPolicy: Never
    backoffLimit: 3
  pollingInterval: 5
  maxReplicaCount: 20
  successfulJobsHistoryLimit: 10
  failedJobsHistoryLimit: 5
  triggers:
    - type: rabbitmq
      metadata:
        host: amqp://user:pass@rabbitmq:5672/
        queueName: images
        queueLength: "1"  # 1 job per message

Multiple Triggers (Composite Scaling)

KEDA uses the maximum across all triggers:

triggers:
  # Scale on queue depth OR high CPU
  - type: rabbitmq
    metadata:
      queueName: orders
      queueLength: "10"
  - type: cpu
    metricType: Utilization
    metadata:
      value: "70"

Scaling Behavior Tuning

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: my-app
spec:
  scaleTargetRef:
    name: my-app
  advanced:
    horizontalPodAutoscalerConfig:
      behavior:
        scaleUp:
          stabilizationWindowSeconds: 0
          policies:
            - type: Percent
              value: 100
              periodSeconds: 15
        scaleDown:
          stabilizationWindowSeconds: 300
          policies:
            - type: Pods
              value: 1
              periodSeconds: 60
  triggers:
    - type: prometheus
      metadata:
        query: sum(rate(http_requests_total[1m]))
        threshold: "50"

Monitoring KEDA

# Check ScaledObject status
kubectl get scaledobjects
kubectl describe scaledobject my-app

# Check generated HPA
kubectl get hpa

# KEDA metrics
kubectl get --raw "/apis/external.metrics.k8s.io/v1beta1" | jq .

# Operator logs
kubectl logs -n keda deploy/keda-operator --tail=50

Prometheus Metrics from KEDA

# KEDA exposes metrics on :8080/metrics
- keda_scaler_active: Is the scaler triggering scale-up?
- keda_scaler_metrics_value: Current metric value
- keda_scaled_object_errors: Error count per ScaledObject

KEDA vs Native HPA

Deepen Your Kubernetes Skills

If you found this article useful, check out my books for hands-on Kubernetes mastery:

• Kubernetes Recipes — A practical guide for container orchestration and deployment with real-world patterns
• Ansible for Kubernetes by Example — Automate Kubernetes cluster operations with Ansible playbooks

Both books follow the same practical, example-driven approach you see in my articles.

Related Content

• Kubernetes HPA Custom Metrics with Prometheus Adapter
• Kubernetes VPA Vertical Pod Autoscaler Guide
• Volcano Batch Scheduling and Gang Scheduling
• Kubernetes CronJob Concurrency Policy Deep Dive