// the find
kedacore/keda
KEDA is a Kubernetes-based Event Driven Autoscaling component. It provides event driven scale for any container running in Kubernetes
KEDA is a CNCF-graduated Kubernetes operator that extends the HPA to scale workloads based on external event sources like Kafka lag, SQS queue depth, or cron schedules — including scaling to zero. It's for platform teams and developers who need more than CPU/memory-based autoscaling in Kubernetes.
- Massive scaler coverage: 60+ built-in scalers covering AWS, Azure, GCP, Kafka, Redis, RabbitMQ, Prometheus, and more — most teams won't need to write a custom scaler
- Clean operator pattern with proper CRDs (ScaledObject, ScaledJob, TriggerAuthentication), admission webhooks for validation, and solid separation between the metrics adapter and the controller
- TriggerAuthentication and ClusterTriggerAuthentication give you a real secrets management story — Vault, Azure Key Vault, GCP Secret Manager all supported without baking credentials into ScaledObject specs
- Actively maintained with nightly e2e tests, OpenSSF scorecard, FOSSA license scanning, and a published support policy — the operational maturity signals are all there
- The go.mod is a mess of replace directives pinning transitive deps because of k8s ecosystem version conflicts — anyone vendoring this or building on top of it will hit dependency hell quickly
- ScaledJob behavior (how it handles in-flight jobs vs. scale-down) is genuinely confusing and the edge cases around job completion are poorly documented; you'll learn the hard way in production
- External scaler interface exists but writing one requires implementing a gRPC service and running it as a separate deployment — the operational overhead is high for what should be a simple extension point
- No built-in support for scaling based on combinations of metrics (e.g., scale when queue depth AND latency are both high); each trigger is evaluated independently and you only get OR semantics between triggers