keda

What is keda?

The keda image runs KEDA, the Kubernetes Event-Driven Autoscaler. KEDA extends the Kubernetes Horizontal Pod Autoscaler (HPA) by allowing workloads to scale up or down based on external events and metrics, not just CPU or memory usage. This means applications can scale in response to queue length in RabbitMQ, Kafka lag, Prometheus metrics, cloud services like Azure Event Hubs or AWS SQS, and dozens of other event sources.

KEDA enables fine-grained, on-demand scaling, including the ability to scale workloads down to zero when idle. This makes it particularly relevant for microservices, event-driven systems, and serverless-style workloads running inside Kubernetes. Its design integrates natively with Kubernetes APIs, ensuring that teams can adopt event-driven scaling without introducing proprietary control planes or additional complexity.

The keda image is widely used in production clusters to optimize resource usage, improve cost efficiency, and deliver responsive scaling for workloads that process asynchronous jobs, streams, or external triggers.

How to use this image

The keda image is normally deployed inside a Kubernetes cluster as a controller. It can be installed with Helm, YAML manifests, or via the OperatorHub ecosystem. Once deployed, scaling is managed through custom resources like ScaledObject and ScaledJob.

‍Deploy with Helm:

<code>helm repo add kedacore https://kedacore.github.io/chartshelm install keda kedacore/keda \  --namespace keda --create-namespace</code>

Example ScaledObject using RabbitMQ as a trigger:

<code>apiVersion: keda.sh/v1alpha1kind: ScaledObjectmetadata:  name: queue-scalerspec:  scaleTargetRef:    name: worker  triggers:  - type: rabbitmq    metadata:      queueName: tasks      host: RabbitMqHost</code>

KEDA writes logs to stdout for collection by Kubernetes logging systems. It does not expose user-facing ports directly; instead, it extends the Kubernetes control plane by responding to external metrics sources.

{{cta}}

Image variants

Published under ghcr.io/kedacore/keda, the image is available in multiple variants:

ghcr.io/kedacore/keda:latest

Tracks the most recent stable release.Use for testing or experimentation.

ghcr.io/kedacore/keda:<version>

Tagged by semantic version (e.g. 2.14.0, 2.13.1).

Recommended for production to ensure compatibility and reproducibility.There are no Alpine, slim, or OS-specific variants. Images are built for standard Kubernetes environments and regularly updated with upstream releases.