Why Is Cloud Run Overriding the Concurrency Setting in Our YAML Deployment?

The issue isn’t coming from setting cpu-throttling to false. Cloud Run expects the maximum concurrent requests parameter to be set in a specific field (containerConcurrency) rather than via an annotation. In your YAML file you’re using the annotation:

YAMLCopied
run.googleapis.com/concurrency: '4'

This annotation isn’t used when deploying a service; Cloud Run will default to 80 concurrent requests per instance unless the concurrency is defined in the proper location in the service configuration. When you change it via the GCP Console, Cloud Run correctly picks up the value because it’s updated in the proper field behind the scenes.

To ensure the concurrency value is applied via your YAML, you should define it in the service spec (using the containerConcurrency field) rather than as an annotation. For example:

YAMLCopied
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: your-service-name
  annotations:
    run.googleapis.com/network-interfaces: '[{"network":"${GCP_VPC}","subnetwork":"${GCP_SUBNET}","tags":["${NETWORK_TAG}"]}]'
    run.googleapis.com/vpc-access-egress: all-traffic
    run.googleapis.com/startup-cpu-boost: 'true'
    autoscaling.knative.dev/minScale: ${MIN_SCALE}
    autoscaling.knative.dev/maxScale: ${MAX_SCALE}
    run.googleapis.com/execution-environment: gen2
    run.googleapis.com/cpu-throttling: 'false'
spec:
  template:
    spec:
      containerConcurrency: 4
      containers:
      - image: <your-image>

Using this approach the concurrency setting will be respected during deployment. Remember that any configuration change creates a new revision, so you should verify that the field is correctly applied to avoid the default 80 concurrency.