KEDA 1.0: Event-Driven Autoscaling for Kubernetes

Introduction

Microsoft and Red Hat jointly announced KEDA 1.0 on November 19, 2019, bringing event-driven autoscaling to any Kubernetes cluster. KEDA plugs directly into the Horizontal Pod Autoscaler to scale deployments based on event streams, queues, and cloud services.

Core Capabilities

• Scaler Catalog: 20+ built-in scalers for Azure Queue, AWS SQS, Kafka, Prometheus, RabbitMQ, Postgres, and more.
• HPA Bridge: KEDA acts as a metrics adapter, feeding event counts into standard HPA calculations.
• Scale-to-Zero: Deployments rest at zero replicas until trigger metrics rise above thresholds.

Extensibility

1. External Scalers: gRPC interface lets platform teams author custom triggers.
2. ScaledObjects & ScaledJobs: CRDs declaratively bind workloads to event sources.
3. Authentication: TriggerAuthentication resources manage cloud credentials and connection strings.

Best Practices

• Combine with Knative or Azure Functions on Kubernetes for full serverless experiences.
• Use ScaledJob for queue workers that should drain messages and exit.
• Monitor via Prometheus metrics exposed by the KEDA operator to watch scaling decisions.

Getting Started

kubectl apply -f https://github.com/kedacore/keda/releases/download/v1.0.0/keda-1.0.0.yaml

Create a ScaledObject for Azure Queue Storage:

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: queue-consumer
spec:
  scaleTargetRef:
    name: worker
  triggers:
  - type: azure-queue
    metadata:
      queueName: orders
      queueLength: "5"

Summary

KEDA 1.0 unlocks event-driven workloads for platform teams that want serverless agility without leaving Kubernetes.