Knative: Operator’s Handbook

Configuring instance count

By default, Knative will scale number of Pods for each Service based on incoming requests:

To adjust these numbers you can specify annotations at spec.template.metadata.annotations.

Note: Make sure annotation values are “quoted”, so they are treated as strings.

apiVersion: serving.knative.dev/v1alpha1
kind: Service
metadata:
  name: hello
spec:
  template:
    metadata:
      annotations:
        autoscaling.knative.dev/minScale: "1"
        # and/or:
        autoscaling.knative.dev/maxScale: "20"
    spec:
      containers:
      - image: gcr.io/google-samples/hello-app:1.0
      [...]

Note that setting minScale > 0 will cause each Revision to always run specified number of Pods despite they're not getting any traffic. In that case, do not forget to clean up old revisions.