CPU allocation

By default, Cloud Run container instances are only allocated CPU during request processing and container startup. You can change this behavior so CPU is always allocated and available even when there are no incoming requests. Setting the CPU to be always allocated can be useful for running background tasks and other asynchronous processing tasks.

Note that even if CPU is always allocated, Cloud Run autoscaling is still in effect, and may terminate container instances if they aren't needed to handle incoming traffic. An instance will never stay idle for more than 15 minutes after processing a request unless it is kept active using minimum instances.

Combining CPU always allocated with a number of minimum instances results in a number of container instances up and running with full access to CPU resources, enabling background processing use cases like pulling Pub/Sub messages.

Pricing impact

If you choose the CPU to be allocated only during request processing, you are charged per request and only when the container instance processes a request. If you choose the CPU always allocated setting, you are charged for the entire lifecycle of the container instance. See the Cloud Run pricing tables for details.

How to choose the appropriate CPU allocation

Choosing the appropriate CPU allocation for your use case depends on several factors, such as traffic patterns, background execution, and cost, each of which is described in the following sections.

Traffic patterns considerations

  • CPU only allocated during request processing is recommended when incoming traffic is sporadic, bursty or spiky.
  • CPU always allocated is recommended when incoming traffic is steady, slowly varying.

Background execution considerations

Selecting CPU always allocated allows you to execute background tasks and other asynchronous processing work after returning responses. For example:

  • Leveraging monitoring agents like OpenTelemetry that may assume to be able to run in the background.
  • Using Go's Goroutines, Node.js async, Java threads, and Kotlin coroutines
  • Using application frameworks that rely on built-in scheduling/background functionalities.

Cost considerations

If you are currently using CPU only allocated during request processing, CPU always allocated is probably more economical if:

  • Your Cloud Run service is processing high number of current requests at a rather steady rate.
  • You do not see a lot of "idle" container instances when looking at the container instance count metric.

You can use the pricing calculator to estimate cost differences.

Setting and updating CPU allocation

Any configuration change leads to the creation of a new revision. Subsequent revisions will also automatically get this configuration setting unless you make explicit updates to change it.

If you are choosing the always-allocated CPU option, you must specify at least 512MiB of memory.

By default, CPU is only allocated during request processing for each container instance. You can change this using the Cloud Console, the gcloud command line, or a YAML file when you create a new service or deploy a new revision:

Console

  1. Go to Cloud Run

  2. Click Create Service if you are configuring a new service you are deploying to. If you are configuring an existing service, click on the service, then click Edit and Deploy New Revision.

  3. If you are configuring a new service, fill out the initial service settings page as desired, then click Next > Advanced settings to reach the service configuration page.

  4. Click the Container tab.

    image

  5. Select the desired CPU allocation under CPU allocation and pricing. Select CPU is only allocated during request processing for your instances to receive CPU only when they are receiving requests. Select CPU is always allocated to to allocate CPU for the entire lifetime of container instances.

  6. Click Create or Deploy.

Command line

You can update the CPU allocation. To set CPUs to be always allocated for a given service:

gcloud beta run services update SERVICE --no-cpu-throttling 

Replace SERVICE with the name of your service.

To set CPU allocation only during request processing:

gcloud beta run services update SERVICE --cpu-throttling 

You can also set CPU allocation during deployment. To set CPUs to be always allocated:

gcloud run deploy --image IMAGE_URL --no-cpu-throttling

To set CPU allocation only during request processing:

gcloud run deploy --image IMAGE_URL --cpu-throttling

Replace IMAGE_URL with a reference to the container image, for example, us-docker.pkg.dev/cloudrun/container/hello:latest.

YAML

You can download and view existing service configuration using the gcloud run services describe --format export command, which yields cleaned results in YAML format. You can then modify the fields described below and upload the modified YAML using the gcloud run services replace command. Make sure you only modify fields as documented.

  1. To view and download the configuration:

    gcloud run services describe SERVICE --format export > service.yaml
  2. Update the cpu attribute:

    apiVersion: serving.knative.dev/v1
    kind: Service
    metadata:
      name: SERVICE
    spec:
      template:
        metadata:
          annotations:
            run.googleapis.com/cpu-throttling: 'BOOLEAN`

    Replace

    • SERVICE with the name of your Cloud Run service
    • BOOLEAN with true to set CPU allocation only during request processing, or false to set CPU to always allocated.
  3. Replace the service with its new configuration using the following command:

    gcloud run services replace service.yaml

Viewing CPU allocation settings

To view the current CPU allocation settings for your service:

Console

  1. Go to Cloud Run

  2. Click the service you are interested in to open the Service details page.

  3. Click the Revisions tab.

  4. In the details panel at the right, the CPU allocation setting is listed under the Container tab.

Command line

  1. Use the following command:

    gcloud run services describe SERVICE
  2. Locate the CPU allocation setting in the returned configuration.