KEDA Learning Roadmap

Goal: From zero event-driven autoscaling experience to production KEDA operations

Learning Phases

Phase 1: KEDA Fundamentals (1 day)

•Why HPA alone can't scale on queue depth/lag — the core problem KEDA solves

•ScaledObject — the CRD that replaces direct HPA management

•Scale-to-zero and activation thresholds

•CNCF graduated project status and ecosystem context

Phase 2: Triggers and Scaling Patterns (2-3 days)

•Kafka lag, RabbitMQ queue depth, cron, Prometheus query triggers

•Combining multiple triggers (scale on whichever is higher)

•ScaledJob vs. ScaledObject — when each fits

Phase 3: Intermediate Configuration (2-3 days)

•TriggerAuthentication — secret-based and cloud-identity-based credential management

•HPA behavior tuning — asymmetric scale-up/scale-down rates

•idleReplicaCount vs. minReplicaCount — the actual scale-to-zero mechanism

•KEDA HTTP Add-on for synchronous HTTP services

Phase 4: Advanced Production Operations (3-4 days)

•KEDA's own metrics-server internals and the External Metrics API chain

•Building a custom External Scaler via gRPC

•Multi-cluster and federation patterns

•KEDA Operator HA and CRD versioning/migration

Phase 5: Troubleshooting and Interview Readiness (ongoing)

•Diagnosing READY: False ScaledObjects and scaling rate limits

•Cold-start latency and cron-timezone bugs

•Practice explaining the HPA-vs-KEDA distinction concretely — Overview's own Kafka-lag example is the canonical answer to "why would you need KEDA at all"

Job Roles This Enables

•Platform Engineer

•Kubernetes/Cloud-Native Infrastructure Engineer

•DevOps Engineer (autoscaling/cost-optimization focus)

Target Certifications

(needs verification — recheck against current source for CNCF's current KEDA-relevant certification offerings, if any)

How to Use This Roadmap

1.Work through phases in order — Phase 3's TriggerAuthentication and HPA-tuning material assumes Phase 2's trigger types are already familiar

2.This technology doesn't have a separate Labs tab — the Fundamentals/Intermediate/Advanced code blocks and a real cluster with a test queue (a local RabbitMQ or Kafka instance works) are the hands-on material

3.Build the portfolio projects in projects.md after Phase 3 — demonstrating a real scale-to-zero-to-N cycle against actual queue depth is far more convincing than describing the configuration

4.Use the Troubleshooting section's real scenarios as active study material — cron-timezone bugs and READY: False states are genuinely common real-world gotchas, not contrived examples

5.Phase 4's custom External Scaler material is worth understanding conceptually even if you never build one — it's the answer to "what if KEDA doesn't have a built-in scaler for my system," a real interview question

Prerequisites

See prerequisites.md in this section for what you should know before starting.

Revision Notes

Total time: 1-2 weeks (part-time)
The idleReplicaCount vs. minReplicaCount distinction is the single
most commonly confused KEDA concept — budget deliberate practice time
there specifically, not just reading comprehension

KEDA — Roadmap