Data Sharing and Snowflake Marketplace · Topic 7 of 10

Clean Rooms - Bringing Users In and Managing Collaborators

Bring users into a Snowflake Data Clean Room, define collaborators and their roles in a collaboration, and manage joins, edits, leaving and teardown.

  • 4 min
In a nutshell
  • Account access and collaboration membership are separate; the installer's account needs an administrator to grant access.
  • Roles: owner (one), data provider, analysis runner; a collaborator can hold several.
  • All collaborators except the owner review before joining.

What it is

Once the clean room app is installed, two different groups of people have to be brought in. The first are users inside your own account who may operate the clean room. The second are collaborators, which are other parties, each with their own Snowflake account, who join a specific collaboration.

Keeping them apart is the point of this page. Account access is a Snowflake privilege question. Collaboration membership is a business agreement written into a specification.

Installed appaccount leveladmin grants accessYour users and rolescan operate clean roomscreate or joinCollaborationspecificationlistsCollaboratorsowner, providers, runners

How it works

Bring users into the clean room

The install covers the whole account, but it does not give anyone access. A clean rooms administrator has to grant access to the users and roles that need it. The usual order is:

  1. The ACCOUNTADMIN who installed the app verifies it (CHECK_MOUNT_STATUS() under SAMOOHA_APP_ROLE).
  2. The administrator grants clean room access to chosen roles, and through them to users. The docs have a managing access topic for the exact roles and grants.
  3. Each user needs a complete user profile (first name, last name, verified email) because the app uses these details.
  4. Users then open the clean room interface in Snowsight, or call the clean room API from a worksheet.

Design guidance for access:

PracticeReason
Grant to roles, not to individual usersJoiners and leavers are handled by role membership
Keep a small group for administrationCreating and tearing down collaborations has consequences for partners
Separate roles for people who only run analysesLeast privilege
Check the user profile dataMissing email or name blocks some operations
Exam tip

Granting access to the clean room does not give anyone access to partner data. Partner data stays behind the collaboration rules, templates and policies.

Define collaborators

A collaborator is an account that takes part in a collaboration. The owner lists collaborators and roles in the collaboration specification when creating it.

RoleResponsibility
OwnerCreates the collaboration, assigns roles, can tear it down. Exactly one; not automatically a runner or provider
Data providerSupplies data offerings to named analysis runners
Analysis runnerRuns permitted templates on permitted data offerings

A collaborator can hold several roles. For example, a retailer can be both data provider and analysis runner, while an advertiser is only an analysis runner.

The specification is the contract. It names the collaborators, the roles, the data offerings, the templates and the policies that apply. A collaborator cannot do anything that the specification does not permit.

Manage collaborators over time

ActionHow it works
CreateThe owner defines the specification; the collaboration becomes visible to everyone listed
JoinEvery collaborator except the owner reviews the collaboration and then joins
Check statusGET_STATUS shows the state of the collaboration and joins
Add or remove a collaboratorThe owner uses EDIT; affected collaborators must approve the change
LeaveA non owner uses LEAVE
EndThe owner uses TEARDOWN to delete the collaboration

Practical rules:

  • Collaborators in other cloud regions need Cross-Cloud Auto-Fulfillment enabled, or they cannot take part.
  • A data provider needs Enterprise Edition or higher; an owner or an analysis runner can be on Standard Edition.
  • Activating results to another Snowflake account requires Enterprise Edition or higher.
  • Only the owner ends a collaboration; a non owner can only leave it.
Ownercreates specificationinvitesCollaboratorsreviewjoinActive collaborationtemplates runEDIT with approvalChanged membershipadd or removeTEARDOWNEndedby owner
Watch out

Removing a data provider takes away the data that provider offered, so analyses that depend on it can no longer run. Tell the runners before you edit.

Audit and governance of collaborators

  • Review the specification before joining. It states what data you will expose and which templates may run on it.
  • Remove collaborators who no longer need access instead of leaving them idle.
  • Keep a record of who owns each collaboration, because the owner is the only one who can tear it down.
  • Rely on the policies in the collaboration to control what your data may be used for.

Common mistakes

Believing installation grants access. An administrator must grant access to users and roles.

Giving the owner the other roles by default. The owner is not automatically a data provider or analysis runner.

Joining without reading the specification. It defines what your data may be used for.

Adding a collaborator in another region without Cross-Cloud Auto-Fulfillment. It does not work.

Expecting a collaborator to leave a collaboration alone to change it. Membership changes use EDIT and need approval.

What to remember for the exam

  • Account access and collaboration membership are separate; the installer's account needs an administrator to grant access.
  • Roles: owner (one), data provider, analysis runner; a collaborator can hold several.
  • All collaborators except the owner review before joining.
  • EDIT changes membership with approval; GET_STATUS, LEAVE and TEARDOWN complete the set.
  • Data provider needs Enterprise Edition or higher; owners and runners can be on Standard.

Check yourself

  • Who can end a collaboration?

    The owner, with TEARDOWN.

  • How does a collaborator in a different cloud region take part?

    Cross-Cloud Auto-Fulfillment must be enabled for them.

  • What must happen before a new collaborator is added through EDIT?

    The affected collaborators must approve the change.

Topic 7 of 10 in Data Sharing and Snowflake Marketplace

Test yourself on Data Sharing and Snowflake Marketplace

The full Snowflake ADA-C02 question bank has 814 questions, with timed practice exams and an explanation for every answer.

This study guide is independent and is not affiliated with, endorsed by, or sponsored by Snowflake. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.