KEDA 2.0: Event-Driven Autoscaling at Scale

Introduction

KEDA 2.0, released on November 19, 2020, accelerates event-driven workloads with revamped APIs, HTTP autoscaling, and richer observability. Version 2.0 cements KEDA as the standard event scaler for Kubernetes.

ScaledObject v2 API

• Consolidates scaling configuration into a stable keda.sh/v1alpha1 schema with clear separation of triggers, cooldowns, and authentication.
• Supports multiple triggers per workload with logical AND/OR semantics to satisfy composite scaling criteria.
• Backwards-compatible: KEDA migrates v1 ScaledObjects automatically during upgrade.

HTTP Add-On & Multi-Metric Scaling

1. HTTP Add-On introduces request queue autoscaling for traditional web services, complementing queue-based scalers.
2. Built-in scalers expand to Redis Streams, Azure Log Analytics, and custom metrics via Prometheus.
3. Global cooldowns and stabilization windows prevent oscillation during bursty traffic.

Operations & Observability

• KEDA metrics server now exports per-trigger activation status, target values, and scaling history.
• Helm chart refactor simplifies upgrades and enables HA operator replicas for large clusters.
• Admission webhooks validate ScaledObjects and authenticate secrets references securely.

Getting Started

kubectl apply -f https://github.com/kedacore/keda/releases/download/v2.0.0/keda-2.0.0.yaml

Define a multi-trigger ScaledObject:

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: orders-worker
spec:
  pollingInterval: 30
  minReplicaCount: 0
  maxReplicaCount: 50
  scaleTargetRef:
    name: worker
  triggers:
  - type: aws-sqs-queue
    metadata:
      queueURL: https://sqs.us-west-2.amazonaws.com/123456789/orders
      queueLength: "10"
  - type: prometheus
    metadata:
      serverAddress: http://prometheus.monitoring:9090
      metricName: http_requests_inflight
      threshold: "100"

Summary

KEDA 2.0 empowers platform teams to deliver serverless-like elasticity for any workload without leaving Kubernetes.