Scales Knative Services in response to request-driven load metrics. The autoscaler is a control-plane component of Knative Serving. It collects per-revision load statistics from the activator and queue-proxies, computes the desired number of pods for each Revision, and drives scaling decisions including scale-to-zero.
The autoscaler is a component of the Knative Serving control plane. It collects per-revision request metrics reported by the activator and the queue-proxy sidecars, aggregates them over stable and panic windows, and computes the desired number of pods for each Revision. Based on those decisions it scales Revisions up and down, including scaling to zero when a Revision receives no traffic.
Use this image when you need a hardened, minimal container image for running the Knative Serving autoscaler as part of a Knative Serving deployment.