Create custom organization policy constraints

This page shows you how to use Organization Policy Service custom constraints to restrict specific operations on the following Google Cloud resources:

• aiplatform.googleapis.com/CustomJob
• aiplatform.googleapis.com/HyperparameterTuningJob
• aiplatform.googleapis.com/NasJob

To learn more about Organization Policy, see Custom organization policies.

About organization policies and constraints

The Google Cloud Organization Policy Service gives you centralized, programmatic control over your organization's resources. As the organization policy administrator, you can define an organization policy, which is a set of restrictions called constraints that apply to Google Cloud resources and descendants of those resources in the Google Cloud resource hierarchy. You can enforce organization policies at the organization, folder, or project level.

Organization Policy provides built-in managed constraints for various Google Cloud services. However, if you want more granular, customizable control over the specific fields that are restricted in your organization policies, you can also create custom constraints and use those custom constraints in an organization policy.

Policy inheritance

By default, organization policies are inherited by the descendants of the resources on which you enforce the policy. For example, if you enforce a policy on a folder, Google Cloud enforces the policy on all projects in the folder. To learn more about this behavior and how to change it, refer to Hierarchy evaluation rules.

Benefits

You can use custom organization policies to allow or deny specific values for Vertex AI training resources. For example, if a request to create a custom training job fails to satisfy custom constraint validation as set by your organization policy, the request fails and an error is returned to the caller.

Limitations

Like all organization policy constraints, policy changes don't apply retroactively to existing resources.

• A new policy doesn't impact existing resource configurations.
• An existing resource configuration remains valid, unless you change a value in its configuration from a compliant to a non-compliant value.

Before you begin

Sign in to your Google Cloud account. If you're new to Google Cloud, create an account to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.

In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

Go to project selector

Make sure that billing is enabled for your Google Cloud project.

Install the Google Cloud CLI.

If you're using an external identity provider (IdP), you must first sign in to the gcloud CLI with your federated identity.

To initialize the gcloud CLI, run the following command:

gcloud init

In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

Go to project selector

Make sure that billing is enabled for your Google Cloud project.

Install the Google Cloud CLI.

If you're using an external identity provider (IdP), you must first sign in to the gcloud CLI with your federated identity.

To initialize the gcloud CLI, run the following command:

gcloud init

1. Ensure that you know your organization ID.

Required roles

To get the permissions that you need to manage custom organization policies, ask your administrator to grant you the Organization Policy Administrator (roles/orgpolicy.policyAdmin) IAM role on the organization resource. For more information about granting roles, see Manage access to projects, folders, and organizations.

You might also be able to get the required permissions through custom roles or other predefined roles.

Create a custom constraint

A custom constraint is defined in a YAML file by the resources, methods, conditions, and actions that are supported by the service on which you are enforcing the organization policy. Conditions for your custom constraints are defined using Common Expression Language (CEL). For more information about how to build conditions in custom constraints using CEL, see the CEL section of Creating and managing custom constraints.

To create a custom constraint, create a YAML file using the following format:

name: organizations/ORGANIZATION_ID/customConstraints/CONSTRAINT_NAME
resourceTypes:
- RESOURCE_NAME
methodTypes:
- CREATE
- UPDATE
condition: "CONDITION"
actionType: ACTION
displayName: DISPLAY_NAME
description: DESCRIPTION

Replace the following:

• ORGANIZATION_ID: your organization ID, such as 123456789.

• CONSTRAINT_NAME: the name you want for your new custom constraint. A custom constraint must start with custom., and can only include uppercase letters, lowercase letters, or numbers. For example, custom.restrictMachineType. The maximum length of this field is 70 characters.

• RESOURCE_NAME: the fully qualified name of the Google Cloud resource containing the object and field you want to restrict. For example, aiplatform.googleapis.com/CustomJob.

• CONDITION: a CEL condition that is written against a representation of a supported service resource. This field has a maximum length of 1000 characters. See Supported resources for more information about the resources available to write conditions against. For example, "resource.jobSpec.workerPoolSpecs.exists(spec, spec.machineSpec.machineType != \"n1-standard-4\")".

• ACTION: the action to take if the condition is met. Possible values are ALLOW and DENY.

• DISPLAY_NAME: a human-friendly name for the constraint. This field has a maximum length of 200 characters.

• DESCRIPTION: a human-friendly description of the constraint to display as an error message when the policy is violated. This field has a maximum length of 2000 characters.

For more information about how to create a custom constraint, see Defining custom constraints.

Set up a custom constraint

After you have created the YAML file for a new custom constraint, you must set it up to make it available for organization policies in your organization. To set up a custom constraint, use the gcloud org-policies set-custom-constraint command:

gcloud org-policies set-custom-constraint CONSTRAINT_PATH

Replace CONSTRAINT_PATH with the full path to your custom constraint file. For example, /home/user/customconstraint.yaml. Once completed, your custom constraints are available as organization policies in your list of Google Cloud organization policies. To verify that the custom constraint exists, use the gcloud org-policies list-custom-constraints command:

gcloud org-policies list-custom-constraints --organization=ORGANIZATION_ID

Replace ORGANIZATION_ID with the ID of your organization resource. For more information, see Viewing organization policies.

Enforce a custom organization policy

You can enforce a constraint by creating an organization policy that references it, and then applying that organization policy to a Google Cloud resource.

Console

1. In the Google Cloud console, go to the Organization policies page.

   Go to Organization policies

2. From the project picker, select the project for which you want to set the organization policy.
3. From the list on the Organization policies page, select your constraint to view the Policy details page for that constraint.
4. To configure the organization policy for this resource, click Manage policy.
5. On the Edit policy page, select Override parent's policy.
6. Click Add a rule.
7. In the Enforcement section, select whether enforcement of this organization policy is on or off.
8. Optional: To make the organization policy conditional on a tag, click Add condition. Note that if you add a conditional rule to an organization policy, you must add at least one unconditional rule or the policy cannot be saved. For more information, see Setting an organization policy with tags.
9. Click Test changes to simulate the effect of the organization policy. Policy simulation isn't available for legacy managed constraints. For more information, see Test organization policy changes with Policy Simulator.
10. To finish and apply the organization policy, click Set policy. The policy requires up to 15 minutes to take effect.

gcloud

To create an organization policy with boolean rules, create a policy YAML file that references the constraint:

      name: projects/PROJECT_ID/policies/CONSTRAINT_NAME
      spec:
        rules:
        - enforce: true
    

Replace the following:

• PROJECT_ID: the project on which you want to enforce your constraint.
• CONSTRAINT_NAME: the name you defined for your custom constraint. For example, custom.restrictMachineType.

To enforce the organization policy containing the constraint, run the following command:

    gcloud org-policies set-policy POLICY_PATH
    

Replace POLICY_PATH with the full path to your organization policy YAML file. The policy requires up to 15 minutes to take effect.

Test the custom organization policy

The following example creates a custom constraint and policy that restricts the machine type.

Before you begin, you must know the following:

• Your organization ID
• A project ID

Create the constraint

1. Save the following file as constraint-custom-job.yaml:

   name: organizations/ORGANIZATION_ID/customConstraints/custom.restrictMachineType
   resourceTypes:
   - aiplatform.googleapis.com/CustomJob
   methodTypes:
   - CREATE
   condition: "resource.jobSpec.workerPoolSpecs.exists(spec, spec.machineSpec.machineType != \"n1-standard-4\")"
   actionType: DENY
   displayName: Restrict machine type custom training jobs
   description: All new custom training jobs must use n1-standard-4 machines.
   

   This defines a constraint where every new custom training job must use the n1-standard-4 machine type. If a custom training job doesn't use this machine type, its creation is denied.

2. Apply the constraint:

   gcloud org-policies set-custom-constraint ~/constraint-custom-job.yaml
   

3. Verify that the constraint exists:

   gcloud org-policies list-custom-constraints --organization=ORGANIZATION_ID
   

   The output is similar to the following:

   CUSTOM_CONSTRAINT                            ACTION_TYPE  METHOD_TYPES   RESOURCE_TYPES                                     DISPLAY_NAME
   custom.restrictMachineType                   DENY         CREATE         aiplatform.googleapis.com/CustomJob                Restrict machine type custom training jobs
   ...
   

Create the policy

1. Save the following file as policy-deny-custom-job.yaml:

   name: projects/PROJECT_ID/policies/custom.restrictMachineType
   spec:
     rules:
     - enforce: true
   

   Replace PROJECT_ID with your project ID.

2. Apply the policy:

   gcloud org-policies set-policy ~/policy-deny-custom-job.yaml
   

3. Verify that the policy exists:

   gcloud org-policies list --project=PROJECT_ID
   

   The output is similar to the following:

   CONSTRAINT                          LIST_POLICY  BOOLEAN_POLICY        ETAG
   custom.restrictMachineType          -            SET                   CLj9zMIGEIiS3K4D-
   

After you apply the policy, wait about two minutes for Google Cloud to start enforcing the policy.

Test the policy

Try to create a Vertex AI custom training job with restricted machine type:

gcloud ai custom-jobs create \
  --region=LOCATION \
  --display-name=JOB_NAME \
  --worker-pool-spec=machine-type=n1-standard-8,replica-count=REPLICA_COUNT,container-image-uri=CUSTOM_CONTAINER_IMAGE_URI

The output is the following:

Operation denied by org policy on resource 'projects/PROJECT_ID/locations/LOCATION': ["customConstraints/custom.restrictMachineType": "All new custom training jobs must use n1-standard-4 machines."]

Example custom organization policies for common use cases

The following table provides the syntax of some custom constraints for common use cases:

Description

Constraint syntax

Restrict machine type for Vertex AI custom training jobs

name: organizations/ORGANIZATION_ID/customConstraints/custom.restrictMachineType resourceTypes: - aiplatform.googleapis.com/CustomJob methodTypes: - CREATE condition: "resource.jobSpec.workerPoolSpecs.exists(spec, spec.machineSpec.machineType != "n1-standard-4")" actionType: DENY displayName: Restrict machine type custom training jobs description: All new custom training jobs must use n1-standard-4 machines.

Vertex AI supported resources

The following table lists the Vertex AI resources that you can reference in custom constraints.

What's next

• Learn more about Organization Policy Service.
• Learn more about how to create and manage organization policies.
• See the full list of managed organization policy constraints.