For the complete documentation index, see llms.txt. This page is also available as Markdown.

Requesting Databricks access

Request just-in-time Unity Catalog privileges, workspace admin, or account admin in Databricks through P0.

The Databricks integration is in beta. Functionality and configuration may change.

Request temporary Unity Catalog privileges, workspace admin, or account admin in your organization's Databricks accounts. When your request is approved, P0 grants the access directly to your existing Databricks user, and removes it when the access expires or is revoked.

This page shows the P0 CLI commands (see Installing p0 CLI to install it). You can also request Databricks access from Slack or the P0 web app, as Requesting access describes. Add --reason to any CLI request to tell approvers why you need the access, and run p0 request databricks --help to see the arguments for your installation.

Before you request

To request access, you must:

  • Exist as a user in the Databricks account you request access in. P0 doesn't create Databricks users.

    • For Unity Catalog privileges, your Databricks username must be your P0 email. P0 grants the privileges to the Databricks user with that username, and doesn't match the other emails on your Databricks user.

    • For workspace admin or account admin, your Databricks username or one of the emails on your Databricks user must match your P0 email.

  • Have a P0 email and a Databricks username in domains that the connector's Allowed email domains match. Your organization sets this when it installs the integration.

On Azure Databricks, your username can be your Microsoft Entra user principal name. If it differs from your P0 email, P0 can't grant you Unity Catalog privileges.

To use Unity Catalog privileges once they're granted, you also need to be assigned to a workspace that can use the catalog, and to have compute there, such as a SQL warehouse or cluster. P0 grants neither. You can request the privileges before you're assigned.

Request Unity Catalog privileges

p0 request databricks privilege <catalog> <privilege>... [--schema <schema>] [--table <table> | --volume <volume> | --function <function>]

P0 grants the privileges you list on the most specific object you name: the catalog, a schema in it, or a table, volume, or function in that schema. Name at most one table, volume, or function, always together with its schema, and give each name without the catalog or schema in front of it.

The catalog must be one your organization added to P0. Give its name, such as main. If two catalogs your organization added share a name, give the one you want as <catalog>@<workspace-id>. A request that gives only the shared name fails with More than one installed catalog is named main. Name it as <catalog>@<workspace ID> instead:, followed by the catalogs to choose from.

For example, to request SELECT and MODIFY on the table main.sales.orders:

To reach an object, Databricks requires USE_CATALOG on its catalog and USE_SCHEMA on its schema. P0 adds both to a request on a table, volume, or function, and USE_CATALOG to a request on a schema. It skips one you already hold, and leaves it in place when your access ends.

P0 adds neither to a request on a catalog, and doesn't add USE_SCHEMA to a request on a schema. To use the objects inside, name those privileges yourself, and leave out any you already hold directly on that catalog or schema: P0 fails a request for a privilege you already have. For example, to read every table in main when you hold neither USE_CATALOG nor USE_SCHEMA on it:

A privilege on a catalog or schema applies to every object in it, including objects created later. Request the narrowest object you need.

P0 grants these privileges: APPLY_TAG, BROWSE, CREATE_FUNCTION, CREATE_MATERIALIZED_VIEW, CREATE_MODEL, CREATE_SCHEMA, CREATE_TABLE, CREATE_VOLUME, EXECUTE, MODIFY, READ_VOLUME, REFRESH, SELECT, USE_CATALOG, USE_SCHEMA, and WRITE_VOLUME. Privilege names aren't case-sensitive. For what each one allows, see Databricks' Unity Catalog privileges reference. P0 grants privileges on Unity Catalog objects only, not on objects in the legacy Hive metastore.

Request workspace admin or account admin

  • workspace-admin makes you an admin of one workspace that your organization added to P0 with workspace-admin requests turned on. Give the workspace's ID, or the name your organization gave it in P0 if no other workspace in the account has that name.

  • account-admin makes you an account admin of the Databricks account, if your organization turned on account-admin requests for it.

  • --account takes the Databricks account ID. You need it only when your organization has added more than one Databricks account.

You don't need to be assigned to a workspace before you request workspace admin on it, as long as you already exist as a user in the workspace's Databricks account.

For example, to become an admin of the workspace named analytics-prod in P0:

If a request fails

P0 checks some things when you submit a request, before anyone approves it. A request fails at submission if:

  • You already have a P0 request that hasn't ended for the same object, or for an admin role on the same workspace or account. The error is P0 already manages your access to <object>. Wait for your previous request to end, or revoke it, then request again. To change your privileges on an object, revoke that request, then ask for every privilege you need in one request.

  • An earlier request of yours for the same object, workspace, or account failed, and may have left its access in place. The error is P0 couldn't finish removing your previous access to <object>. Ask an admin to fix it in Databricks and finish the revoke in P0, then request again.

  • You already hold one of the privileges you named, directly on that object, or the admin role you asked for, and P0 didn't grant it. The error is This principal already has this access. Leave that privilege out of the request.

  • Your P0 email or Databricks username isn't in an allowed domain. The error starts with Domain verification refused. Ask your P0 admin which domains the connector allows.

  • The connector's Allowed email domains setting is missing or invalid. The error starts with Domain verification is not configured. Ask your P0 admin to fix it.

  • You asked for an admin role that your organization hasn't turned on for that account or workspace. The error starts with Account-admin requests aren't turned on or Workspace-admin requests aren't turned on, and names the setting. Ask your P0 admin to turn it on.

  • The workspace you named isn't one your organization added to P0 for that account. The error starts with No workspace <workspace> is installed in P0. Check the ID or name, or ask your P0 admin to add the workspace.

  • More than one workspace your organization added to the account has the name you gave. The error starts with More than one workspace installed for Databricks account <account-id> is named <workspace> and lists their IDs. Give the ID instead.

  • For an admin request, more than one Databricks user matches your P0 email. The error starts with More than one Databricks user matches and lists them. Ask your Databricks admin to remove the duplicate.

If your organization turns off an admin setting, or removes the workspace from P0, after you submit a request for that role, the request fails the same way once it's approved and P0 goes to grant it.

If no Databricks user has your P0 email as its username, a request for Unity Catalog privileges doesn't fail until it's approved and P0 goes to grant the access. The error is No Databricks user `<your-email>`. P0 grants Unity Catalog privileges only to the Databricks user whose username is this email. Ask a Databricks admin to add that user, then request again.

If no Databricks user matches your P0 email, an admin request doesn't fail until it's approved and P0 goes to grant the access. The error is No Databricks user `<your-email>`. Users must already exist. Ask your Databricks admin to add you to the Databricks account, then request again.

Once P0 grants the access, it reports the grant only after Databricks shows the new access, and the revoke only after Databricks shows it removed.

Last updated