Dataproc permissions and IAM roles

Overview

Google Cloud Identity and Access Management (IAM) allows you to control user and group access to your project's resources. This document focuses on the IAM permissions relevant to Dataproc and the IAM roles that grant those permissions.

Dataproc Permissions

Dataproc permissions allow users to perform specific actions on Dataproc clusters, jobs, operations, and workflow templates. For example, the dataproc.clusters.create permission allows a user to create Dataproc clusters in your project. You don't directly give users permissions; instead, you grant them roles, which have one or more permissions bundled within them.

The following tables list the permissions necessary to call Dataproc APIs (methods). The tables are organized according to the APIs associated with each Dataproc resource (clusters, jobs, operations, and workflow templates).

Clusters Permissions

Method Required Permission(s)
projects.regions.clusters.create 1, 2 dataproc.clusters.create
projects.regions.clusters.get dataproc.clusters.get
projects.regions.clusters.list dataproc.clusters.list
projects.regions.clusters.patch 1, 2, 3 dataproc.clusters.update
projects.regions.clusters.delete 1 dataproc.clusters.delete
projects.regions.clusters.getIamPolicy dataproc.clusters.getIamPolicy
projects.regions.clusters.setIamPolicy dataproc.clusters.setIamPolicy

Notes:

  1. The dataproc.operations.get permission is also required to get status updates from gcloud command-line tool.
  2. The dataproc.clusters.get permission is also required to get the result of the operation from gcloud command-line tool.
  3. dataproc.autoscalingPolicies.use permission is also required to enable an autoscaling policy on a cluster.

Jobs Permissions

Method Required Permission(s)
projects.regions.jobs.submit 1, 2 dataproc.jobs.create
dataproc.clusters.use
projects.regions.jobs.get dataproc.jobs.get
projects.regions.jobs.list dataproc.jobs.list
projects.regions.jobs.cancel 1 dataproc.jobs.cancel
projects.regions.jobs.patch 1 dataproc.jobs.update
projects.regions.jobs.delete 1 dataproc.jobs.delete
projects.regions.jobs.getIamPolicy dataproc.jobs.getIamPolicy
projects.regions.jobs.setIamPolicy dataproc.jobs.setIamPolicy

Notes:

  1. The gcloud command-line tool additionally requires dataproc.jobs.get in order for the jobs submit, jobs wait, jobs update, jobs delete, and jobs kill commands to function properly.

  2. The gcloud command-line tool additionally requires dataproc.clusters.get permission to submit jobs. For an example of setting the permissions necessary for a user to run gcloud dataproc jobs submit on a specific cluster using Dataproc Granular IAM, see Submitting Jobs with Granular IAM.

Operations Permissions

Method Required Permission(s)
projects.regions.operations.get dataproc.operations.get
projects.regions.operations.list dataproc.operations.list
projects.regions.operations.cancel dataproc.operations.cancel
projects.regions.operations.delete dataproc.operations.delete
projects.regions.operations.getIamPolicy dataproc.operations.getIamPolicy
projects.regions.operations.setIamPolicy dataproc.operations.setIamPolicy

Workflow Template Permissions

Method Required Permission(s)
projects.regions.workflowTemplates.instantiate dataproc.workflowTemplates.instantiate
projects.regions.workflowTemplates.instantiateInline dataproc.workflowTemplates.instantiateInline
projects.regions.workflowTemplates.create dataproc.workflowTemplates.create
projects.regions.workflowTemplates.get dataproc.workflowTemplates.get
projects.regions.workflowTemplates.list dataproc.workflowTemplates.list
projects.regions.workflowTemplates.update dataproc.workflowTemplates.update
projects.regions.workflowTemplates.delete dataproc.workflowTemplates.delete
projects.regions.workflowTemplates.getIamPolicy dataproc.workflowTemplates.getIamPolicy
projects.regions.workflowTemplates.setIamPolicy dataproc.workflowTemplates.setIamPolicy

Notes:

  1. Workflow Template permissions are independent of Cluster and Job permissions. A user without create cluster or submit job permissions may create and instantiate a Workflow Template.

  2. The gcloud command-line tool additionally requires dataproc.operations.get permission to poll for workflow completion.

  3. The dataproc.operations.cancel permission is required to cancel a running workflow.

Autoscaling Policies Permissions

Method Required Permission(s)
projects.regions.autoscalingPolicies.create dataproc.autoscalingPolicies.create
projects.regions.autoscalingPolicies.get dataproc.autoscalingPolicies.get
projects.regions.autoscalingPolicies.list dataproc.autoscalingPolicies.list
projects.regions.autoscalingPolicies.update dataproc.autoscalingPolicies.update
projects.regions.autoscalingPolicies.delete dataproc.autoscalingPolicies.delete
projects.regions.autoscalingPolicies.getIamPolicy dataproc.autoscalingPolicies.getIamPolicy
projects.regions.autoscalingPolicies.setIamPolicy dataproc.autoscalingPolicies.setIamPolicy

Notes:

  1. dataproc.autoscalingPolicies.use permission is required to enable an autoscaling policy on a cluster with a clusters.patch method request.

Dataproc Roles

Identity and Access Management (IAM) Dataproc roles are a bundle of one or more permissions. You grant roles to users or groups to allow them to perform actions on the Dataproc resources in your project. For example, the Dataproc Viewer role contains the dataproc.*.get and dataproc.*.list permissions, which allow a user to get and list Dataproc clusters, jobs, and operations in a project.

The following table lists the Dataproc IAM roles and the permissions associated with each role:

Role ID Permissions
roles/dataproc.admin dataproc.*.getIamPolicy
dataproc.*.setIamPolicy
dataproc.*.create
dataproc.*.get
dataproc.*.list
dataproc.*.delete
dataproc.*.update
dataproc.clusters.use
dataproc.jobs.cancel
dataproc.workflowTemplates.instantiate
dataproc.workflowTemplates.instantiateInline
compute.machineTypes.get
compute.machineTypes.list
compute.networks.get
compute.networks.list
compute.projects.get
compute.regions.get
compute.regions.list
compute.zones.get
compute.zones.list
resourcemanager.projects.get
resourcemanager.projects.list
roles/dataproc.editor dataproc.*.create
dataproc.*.get
dataproc.*.list
dataproc.*.delete
dataproc.*.update
dataproc.clusters.use
dataproc.jobs.cancel
dataproc.workflowTemplates.instantiate
dataproc.workflowTemplates.instantiateInline
compute.machineTypes.get
compute.machineTypes.list
compute.networks.get
compute.networks.list
compute.projects.get
compute.regions.get
compute.regions.list
compute.zones.get
compute.zones.list
resourcemanager.projects.get
resourcemanager.projects.list
roles/dataproc.viewer dataproc.*.get
dataproc.*.list
compute.machineTypes.get
compute.regions.get
compute.regions.list
compute.zones.get
resourcemanager.projects.get
resourcemanager.projects.list
roles/dataproc.worker (for service accounts only) dataproc.agents.*
dataproc.tasks.*
logging.logEntries.create
monitoring.metricDescriptors.create
monitoring.metricDescriptors.get
monitoring.metricDescriptors.list
monitoring.monitoredResourceDescriptors.get
monitoring.monitoredResourceDescriptors.list
monitoring.timeSeries.create
storage.buckets.get
storage.objects.create
storage.objects.get
storage.objects.list
storage.objects.update
storage.objects.delete
storage.objects.getIamPolicy
storage.objects.setIamPolicy

Notes:

  • "*" signifies "clusters," "jobs," or "operations," except the only permissions associated with dataproc.operations. are get, list, and delete.
  • The compute permissions listed above are needed or recommended to create and view Dataproc clusters when using the Google Cloud Console or the Cloud SDK gcloud command-line tool.
  • To allow a user to upload files, grant the Storage Object Creator role. To allow a user to view job output, grant the Storage Object Viewer role. Note that granting either of these Storage roles gives the user the ability to access any bucket in the project.
  • A user must have monitoring.timeSeries.list permission in order to view graphs on the Google Cloud Console→Dataproc→Cluster details Overview tab.
  • A user must have compute.instances.list permission in order to view instance status and the master instance SSH menu on the Google Cloud Console→Dataproc→Cluster details VM Instances tab. For information on Google Compute Engine roles, see Compute Engine→Available IAM roles).
  • To create a cluster with a user-specified service account, the specified service account must have all permissions granted by the Dataproc Worker role. Additional roles may be required depending on configured features. See Service Accounts for list of additional roles.

Project Roles

You can also set permissions at the project level by using the IAM Project roles. Here is a summary of the permissions associated with IAM Project roles:

Project Role Permissions
Project Viewer All project permissions for read-only actions that preserve state (get, list)
Project Editor All Project Viewer permissions plus all project permissions for actions that modify state (create, delete, update, use, cancel)
Project Owner All Project Editor permissions plus permissions to manage access control for the project (get/set IamPolicy) and to set up project billing

IAM Roles and Dataproc Operations Summary

The following table summarizes the Dataproc operations available based on the role granted to the user, with caveats noted.

Operation Project Editor Project Viewer Dataproc Admin Dataproc Editor Dataproc Viewer
Get/Set Dataproc IAM permissions No No Yes No No
Create cluster Yes No Yes Yes No
List clusters Yes Yes Yes Yes Yes
Get cluster details Yes Yes Yes 1, 2 Yes 1, 2 Yes 1, 2
Update cluster Yes No Yes Yes No
Delete cluster Yes No Yes Yes No
Submit job Yes No Yes 3 Yes 3 No
List jobs Yes Yes Yes Yes Yes
Get job details Yes Yes Yes 4 Yes 4 Yes 4
Cancel job Yes No Yes Yes No
Delete job Yes No Yes Yes No
List operations Yes Yes Yes Yes Yes
Get operation details Yes Yes Yes Yes Yes
Delete operation Yes No Yes Yes No

Notes:

  1. The performance graph is not available unless the user also has a role with the monitoring.timeSeries.list permission.
  2. The list of VMs in the cluster will not include status information or an SSH link for the master instance unless the user also has a role with the compute.instances.list permission.
  3. Jobs that include files to be uploaded cannot be submitted unless the user also has the Storage Object Creator role or has been granted write access to the staging bucket for the project.
  4. Job output is not available unless the user also has the Storage Object Viewer role or has been granted read access to the staging bucket for the project.

Service accounts

When you call Dataproc APIs to perform actions in a project where your cluster is located, such as creating VM instances within the project, Dataproc performs these actions on your behalf by using a service account that has the permissions required to perform the actions. The service accounts listed below have the permissions required to perform Dataproc actions in the project where your cluster is located:

  • Dataproc first attempts to use service-[project-number]@dataproc-accounts.iam.gserviceaccount.com.
  • If that service account doesn't exist, Dataproc will fall back to using the Google APIs service account, [project-number]@cloudservices.gserviceaccount.com.

If the cluster uses a Shared VPC network, a Shared VPC Admin must grant both of the above service accounts the role of Network User for the Shared VPC host project. For more information, refer to:

IAM management

You can get and set IAM policies using the Google Cloud Console, the IAM API, or the gcloud command-line tool.

What's next