What’s New in KEDA 2.0? Expanded Triggers and ScaledJob Support

Overview

KEDA (Kubernetes Event‑Driven Autoscaling) 2.0 has been released as a CNCF sandbox project. It extends the native Horizontal Pod Autoscaler (HPA) by allowing scaling decisions to be driven by external event sources such as Azure, AWS, GCP, Redis, Kafka, Prometheus, and many others.

Architecture

KEDA is a lightweight, single‑purpose component that runs alongside standard Kubernetes resources. It does not replace HPA but complements it, enabling event‑driven scaling for selected workloads while leaving other applications unaffected.

Custom Resources

In KEDA 1.x a single ScaledObject could target either a Deployment or a Job. KEDA 2.0 introduces a separate ScaledJob custom resource, allowing independent autoscaling of batch jobs. Both resources can now reference multiple triggers.

Scaling Logic

When multiple triggers are attached to a ScaledObject or ScaledJob, each scaler reports a desired replica count. KEDA selects the highest replica count among all triggers and applies it to the target workload.

New Scalers in 2.0

Azure Log Analytics

IBM MQ

Native CPU scaler (replaces the need to combine HPA with ScaledObjects)

External push‑model scaler for custom scaling logic

A Metrics API extender has also been added, allowing KEDA to consume metrics exposed via REST APIs (e.g., internal services or Microsoft Dynamics CRM) without writing a dedicated scaler.

Community and Governance

The project has attracted contributions from IBM, Pivotal, VMware, Astronomer, and others, and collaborates with the Knative project. To ensure vendor‑neutral governance, the original maintainers donated KEDA to the CNCF, where it currently resides as a sandbox project with plans to graduate to incubation.

References

https://www.ithome.com.tw/news/140966

https://keda.sh/blog/keda-2.0-release/