Event sources and built-in KEDA scalers

Kedify’s KEDA distribution uses KEDA’s event-source model for queues, streams, databases, schedules and external metrics. Choose the source that reflects actual demand, then use the scaler and authentication contract for your installed KEDA minor release.

Choose replicas or jobs

A ScaledObject changes replicas of a long-lived worker/service. A ScaledJob creates finite jobs with separate completion, retry and concurrency semantics. Use the upstream scaler and authentication references below for your event source.

Scroll to explore

Diagram description

A queue or event source can feed either scaling model. With ScaledObject, KEDA activates persistent workers and HPA or KPA controls their active replica count. With ScaledJob, KEDA creates Kubernetes Jobs whose pods process work and complete. Acknowledgement, retries, visibility and idempotency still belong to the queue and application.

Versioned upstream references

These links open the latest KEDA documentation; select your installed minor version before copying fields.

• Scaler catalog: source-specific signals, limits and prerequisites.
• ScaledObject and ScaledJob: desired-state contracts.
• Authentication: secret and identity options; avoid embedding credentials in trigger metadata.
• Cron scaler: desired replicas during a known window. ScalingPolicy instead changes scaling configuration fields.

Verify the source and workload

Check that the source is reachable and its metric has the intended units, labels and freshness. Use authentication with the least necessary source permissions. Observe activation, desired replicas/jobs, ready capacity and completed work. Test scale-in with in-flight work before production use; retries need application-level idempotency.

Horizontal scaling compares queue, HTTP and OTel signals and explains the control loop. Horizontal tuning covers fallback and stabilization; workload diagnostics starts at the failing layer.