Cloud Run Quotas and Limits

This page contains usage quota and limits that apply when using Cloud Run.

The number of Cloud Run resources is limited. Quotas for Cloud Run encompass API rate limits, which affect the rate at which you can call the Cloud Run Admin API .

There is no direct limit for:

  • The size of container images you can deploy.
  • The number of concurrent requests served by a Cloud Run service .

Resource limits for Cloud Run

To go beyond limits that can be increased, request a quota increase . To go beyond per project limits that cannot be increased, create new resources in a different Google Cloud project or region.

Resource Scope Description Limit Can be increased
Service
 per project and region Maximum number of services 1000 No
Job
 per project and region Maximum number of jobs 1000 No
Job execution
 per project and region Maximum number of running Job executions 1000 No
Revision
 per service Maximum number of revisions per service. When limit is reached, non-serving revisions are automatically deleted in historical order 1000 No
Revision
 per project and region Maximum number of revisions serving traffic 4000 No
Revision tag
 per project and region Maximum number of revision tags. When the number of revision tags reaches this limit, tags that don't have a traffic percentage are automatically deleted in historical order. 2000 No
Job execution
 per job Retention limit for completed job executions. When the number of completed executions for a job reaches this limit, executions are automatically deleted in historical order 1,000 No
Job execution task 1
per job execution Maximum number of tasks running in parallel Depends on selected region and CPU and memory configurations. This limit might be greater in high-capacity regions or lower in recently opened regions. Yes
Job execution task
 per job execution Maximum tasks timeout value 24 hours No
Job execution task
 per job execution Maximum number of tasks in a single job 10,000 No
Job execution task
 per job execution Maximum number of task retries in a job 10 No
Container instance 1
per revision Maximum number of container instances Depends on selected region and CPU and memory configurations . This limit might be greater in high-capacity regions or lower in recently opened regions. Yes
Container instance 1
per project and region Maximum number of container instances of all running job executions Depends on selected region and CPU and memory configurations. This limit might be greater in high-capacity regions or lower in recently opened regions. Yes
Container instance 2
per revision and region Maximum number of container instances using Direct VPC egress 100-200, depending on selected region configurations. Yes
Container instance
 per container instance Startup timeout, in minutes 4 No
Memory
 per container instance Maximum memory size, in GiB 32 No
CPU
 per container instance Maximum number of vCPU 8 No
CPU
 per project and region Maximum total CPU, in milli vCPU, allocated across all instances over a 1 minute period. Depends on selected region. This limit might be greater in high-capacity regions or lower in recently opened regions. Yes
Memory
 per project and region Maximum total memory, in bytes, allocated across all instances over a 1 minute period. Depends on selected region. This limit might be greater in high-capacity regions or lower in recently opened regions. Yes
GPU instance 3
per project per region Maximum number of container instances with GPU 0 3 Yes
Disk
 per container instance Maximum writable, in-memory filesystem , limited by instance memory, in GiB 32 No
Environment variable
 per variable Maximum variable length, in Kb 32 No
Domain mapping
 per top domain and per week Maximum number of SSL certificates 50 No
Domain mapping
 per top domain and per week Maximum number of duplicate SSL certificates 5 No
Access token
 per container instance per second Maximum number of unique access token generated 50 No
Identity token
 per container instance per second Maximum number of unique identity token generated 50 No
Files opened
 per container instance Maximum number of files that can be opened at the same time. Corresponds to
/proc/sys/fs/file-max .		 25000 No

1 This regional quota is used in a few cases:

  1. It controls the maximum value that can be picked for the maximum instance attribute of a revision. Once it is granted in a given region, all revisions in that region can go up to the granted limit.

  2. It controls the maximum parallelism of a job. Once it is granted in a given region, all jobs in that region can go up to the granted limit.

  3. It controls the total container instances for running job executions in a region.

2 This regional quota is used in a few cases:

  1. It controls the maximum value that can be picked for the maximum instance attribute of a revision using Direct VPC egress. Once it is granted in a given region, all revisions in that region can go up to the granted limit.

  2. It controls the maximum parallelism of a job using Direct VPC egress. Once it is granted in a given region, all jobs in that region can go up to the granted limit.

3 To access GPU, you must request a quota increase for Total Nvidia L4 GPU allocation, per project per region .

Networking limits for Cloud Run

Request limits for Cloud Run

Description Limit Notes
Maximum number of concurrent requests per instance
 1000
Maximum number of concurrent streams per HTTP/2 client connection
 100
Maximum time before timeout per request
 60 minutes
Maximum HTTP/1 request size
 32 MiB if using HTTP/1 server. No limit if using HTTP/2 server.
Maximum HTTP/1 response size
 32 MiB if not using Transfer-Encoding: chunked or streaming mechanisms
Outbound connections per second per instance
 700
Outbound DNS resolutions per second per instance
 1000
Inbound requests per second to an HTTP/1 container port per instance
 800 Doesn't apply to HTTP/2 container ports.

Bandwidth limits for Cloud Run

The following bandwidth limits apply to Cloud Run instances:

Description Limit Notes
Maximum bits per instance for egress over Direct VPC
1 Gbps Egress over Direct VPC egress to destinations on the VPC network.
Maximum bits per instance, excluding egress over Direct VPC
 600 Mbps Based on the sum of ingress and egress bits, excluding egress over Direct VPC egress to destinations on the VPC network.
Maximum total packet rate per instance, excluding egress over Direct VPC
 64,000 packets per second Based on the sum of ingress packets and egress packets, excluding egress over Direct VPC egress to destinations on the VPC network.

If either of the two limits is reached, the Cloud Run instance will have limited bandwidth.

Cloud Run Admin API rate Limits

The following rate limits apply to the Cloud Run Admin API . They do not apply to the requests reaching your deployed Cloud Run services.

Quota Description Limit Can be increased Scope
Cloud Run Admin API read requests
The number of API reads per 60 seconds per region. This is not the number of read requests to your Cloud Run services, which is not limited. 3,000 per 60 seconds Yes per project and region
Cloud Run Admin API write requests
The number of API writes per 60 seconds per region. This is not the number of write requests to your Cloud Run services, which is not limited. 180 per 60 seconds Yes per project and region
Job Run
Maximum number of times a job can be executed per 60 seconds per region. 180 per 60 seconds Yes per project and region

How to increase quota

To request a higher quota value using the Google Cloud console:

1. Go to the Quotas & System Limitspage:

Go to Quotas & System Limits

  1. Find the quota you want to increase in the Quotacolumn. Use the Filtersearch box to search for your quota.

  2. Select the checkbox to the left of your quota.

  3. Click Edit . The Quota changesform displays.

  4. In the Quota changesform, enter the increased quota that you want for your project in the New limitfield.

  5. Complete any additional fields in the form, and then click Done.

  6. Click Submit request.

Although the previous procedure applies to most quota increase requests, you might encounter one of the following exceptions:

  • In the case where there is an existing usage cap the usage cap must be removed before the quota can be increased. You will be presented with a disclosure and agreement. The disclosure states that the usage cap will be deleted and the limit will be updated to the default limit immediately. Your quota increase request will then be processed by Google Cloud normally.

  • Sometimes, the Google Cloud console redirects you to a separate form to request an increased limit. After you submit the form, Google Cloud acknowledges your request by email.

  • Some quotas cannot be updated using the Google Cloud console. If you find that you cannot change a quota from the console, request the increase from Cloud Customer Care . The Billing team does not handle quota adjustments.

Google recommends that you create a different quota increase request for each class of resources. For example, you should separate the per-project network quota increases from the non-networking Compute Engine quota increases. If different classes of requests are combined, one increase request can delay the batch if its approval requires more evaluation time.

If your quota increase request requires approval, Google Cloud sends an email acknowledging receipt of your request. If you need further assistance, you can respond to the email. After reviewing your request, Cloud Customer Care sends you a second email notifying you whether the quota increase was approved or denied. If applicable, the email provides the effective date of the increase.

To find out more about how the quota increase process works, see About quota increase requests .