Cloud Dataflow access control guide

Overview

To limit access for users within a project or organization, you can use Identity and Access Management (IAM) roles for Dataflow. You can restrict access to Dataflow-related resources, as opposed to granting users viewer, editor, or owner access to the entire Google Cloud project.

This page focuses on how to use Dataflow's IAM roles. For a detailed description of IAM and its features, see the IAM documentation.

Every Dataflow method requires the caller to have the necessary permissions. For a list of the permissions and roles Dataflow supports, see the following section.

Permissions and roles

This section summarizes the permissions and roles Dataflow IAM supports.

Required permissions

The following table lists the permissions that the caller must have to call each method:

Method	Required Permissions
dataflow.jobs.create	dataflow.jobs.create
dataflow.jobs.cancel	dataflow.jobs.cancel
dataflow.jobs.updateContents	dataflow.jobs.updateContents
dataflow.jobs.list	dataflow.jobs.list
dataflow.jobs.get	dataflow.jobs.get
dataflow.messages.list	dataflow.messages.list
dataflow.metrics.get	dataflow.metrics.get
dataflow.jobs.snapshot	dataflow.jobs.snapshot

Roles

The following table lists the Dataflow IAM roles with a corresponding list of all the permissions each role includes. Every permission is applicable to a particular resource type.

Role	Title	Description	Permissions	Lowest resource
roles/dataflow.admin	Dataflow Admin	Minimal role for creating and managing dataflow jobs.	compute.machineTypes.get dataflow.* resourcemanager.projects.get resourcemanager.projects.list storage.buckets.get storage.objects.create storage.objects.get storage.objects.list
roles/dataflow.developer	Dataflow Developer	Provides the permissions necessary to execute and manipulate Dataflow jobs.	dataflow.* resourcemanager.projects.get resourcemanager.projects.list	Project
roles/dataflow.viewer	Dataflow Viewer	Provides read-only access to all Dataflow-related resources.	dataflow.jobs.get dataflow.jobs.list dataflow.messages.* dataflow.metrics.* dataflow.snapshots.get dataflow.snapshots.list resourcemanager.projects.get resourcemanager.projects.list	Project
roles/dataflow.worker	Dataflow Worker	Provides the permissions necessary for a Compute Engine service account to execute work units for a Dataflow pipeline.	compute.instanceGroupManagers.update compute.instances.delete compute.instances.setDiskAutoDelete dataflow.jobs.get logging.logEntries.create storage.objects.create storage.objects.get	Project

The Dataflow Worker role (roles/dataflow.worker) provides the permissions (dataflow.workItems.lease, dataflow.workItems.update, and dataflow.workItems.sendMessage) necessary for a Compute Engine service account to execute work units for an Apache Beam pipeline. It should typically only be assigned to such an account, and only includes the ability to request and update work from the Dataflow service.

The Dataflow Service Agent role (roles/dataflow.serviceAgent) provides a Dataflow service account access to managed resources. This Service Agent role is assigned to a Dataflow service account when you create a Dataflow project.

Creating jobs

To a create a job, roles/dataflow.admin includes the minimal set of permissions required to run and examine jobs.

Alternatively, the following permissions are required:

• The roles/dataflow.developer role, to instantiate the job itself.
• The roles/compute.viewer role, to access machine type information and view other settings.
• The roles/storage.objectAdmin role, to provide permission to stage files on Cloud Storage.

Example role assignment

To illustrate the utility of the different Dataflow roles, consider the following breakdown:

• The developer who creates and examines jobs needs the roles/dataflow.admin role.
• For more sophisticated permissions management, the developer interacting with the Dataflow job needs the roles/dataflow.developer role.
  • They need the roles/storage.objectAdmin or a related role to stage the required files.
  • For debugging and quota checking, they need the project roles/compute.viewer role.
  • Absent other role assignments, this role allows the developer to create and cancel Dataflow jobs, but not interact with the individual VMs or access other Cloud services.
• The controller service account needs the roles/dataflow.worker role to process data for the Dataflow service. To access job data, the service accounts needs other roles such as roles/storage.objectAdmin.

Assigning Dataflow roles

Dataflow roles can currently be set on organizations and projects only.

To manage roles at the organizational level, see Access Control for Organizations Using IAM.

To set project-level roles, see Granting, changing, and revoking access to resources.