This page covers permissions for databases and tables. If you haven’t already, check out our Permissions overview.
Now that you have some groups, you’ll want to control their data access by going to Admin settings > Permissions. You’ll see an interactive table that displays all of your databases and all of your groups, and the level of access your groups have for each database.
Setting permissions on a database
You can set various levels of permissions on a data source, from querying access to managing the database connection.
* Available on paid plans.
Data access levels determine which data people can use to ask new questions. Data access is distinct from collection permissions, which determine which existing things people can view: dashboards, questions, and models. Metabase provides both blunt and sharp tools for you to set up data permissions that suit your needs.
You can click on any cell in the permissions table to change a group’s access level. When you’re done making your changes, just click the Save changes button in the top-right, and you’ll see a confirmation dialog summarizing the changes.
Metabase provides different types of data access:
- Unrestricted (including native/SQL editing access)
- Granular (which includes Sandboxed access)
- No self-service
Members of the group can create questions using the graphical query builder on data from all tables (within all namespaces/schemas, if your database uses those), including any tables that might get added to this database in the future.
Granular access allows administrators to explicitly set data access to tables or schemas within a database, with “data access” here meaning the ability to create questions using the graphical query builder. In practice, this means that:
- Admins can set the group’s access to individual tables to either Unrestricted, No self-service, or Sandboxed access.
- If a new table gets added to this database in the future, the group won’t get access to that new table. An administrator would need to explicitly grant access to that table.
Note that Block access is unavailable for individual tables/schemas. Block is a database-level setting; you can only block the entire database.
No self-service access
No self-service prevents people in a group from using the graphical query builder to create new questions that query that database, or from seeing this database in the Browse Data section of your Metabase. Groups with No self-service access can still see saved questions that query this data if they 1) have access to the appropriate collection, and 2) aren’t in a group with blocked access to the database.
For now, impersonation access is only available for PostgreSQL and Snowflake.
Impersonation access allows you to associate user attributes with database-defined roles and their privileges. Metabase queries made by people with attributes that you define will respect the grants given to the database roles.
You can use impersonation to give people access to the native/SQL editor, while at the same time restricting their access to data based on a specific database role. And not just table-level access, but row-level access—or however you define access for that role in your database. Effectively what this means is that you can use impersonation to set up data sandbox-like access to your data, while letting people use the SQL editor to query that data. The difference is that, instead of setting up a data sandbox in Metabase, you need to set up that row-level security via the privileges granted to a role in your database.
When you connect Metabase to a database, Metabase uses the database user account you provided that has one or more database roles. When you give a group in Metabase unrestricted access to a database, that group will have the same privileges as the user account that you used to connect Metabase to that database.
If instead you want to give a group SQL access to some, but not all, of the schemas or tables in that database, you can create an additional role in your database that only includes a subset of those tables—or even specific row-level access—and then use Metabase’s impersonation feature to associate a user attribute with that role. Essentially what Metabase will do is take the user attribute and pass that attribute as a string into a
SET ROLE or
USE ROLE command for the database before Metabase executes the query.
Connection impersonation does not apply to users in the Metabase admins group, as their more permissive privileges take precedence.
Setting up connection impersonation
In your database:
- Create a new role.
- Grant that role privileges.
For exactly how to create a new role in your database and grant that role privileges, you’ll need to consult your database’s documentation. We also have some docs on users, roles, and privileges that can help you get started.
In your Metabase:
- Create a new group, or select an existing group.
- Assign a user attribute to people in that group. You’ll use this user attribute to associate people in that group with a role that you created in your database. For example, if you created a role named “Sales” in your database with access to a subset of tables, you would add a user attribute “Sales” to the group. The user attribute should match the name of the role in your database. Only some databases enforce case sensitivity, so you might want to make sure the attribute name and role match exactly just in case.
- Next, you’ll need to apply the impersonation access to that group. Go to Admin settings > Permissions > Data.
- Select the database you want to set permissions on.
- Find the group that you want to associate with the database role you created. Under Data access for that group, select Impersonation.
- From the dropdown, select the user attribute that you added that maps to the role you want the group to use when querying the database.
- Save your changes.
Keep in mind that Metabase gives people the most permissive access to data across all of their groups. So if a person is in one group with impersonated access, and one group with unrestricted access, the unrestricted access would override the impersonated access.
Block ensures people in a group can’t see the data from this database, regardless of their permissions at the collection level.
Even if a question is in a collection that the group has access to, but that question queries a database that is blocked for that group, people in that group won’t be able to view that question unless they’re in another group with the relevant data permissions. Essentially, what Block does is make collections permissions insufficient to view a question.
If a person in that blocked group belongs to another group that does have the corresponding data access, that more privileged access will take precedence (overruling the block), and they’ll be able to view that question.
“Corresponding data access” here refers to whether the saved question was created using the graphical query builder, or the native/SQL editor, as the required permissions to overrule a block differ depending on how the question was created.
- If the question was created using the graphical query builder, the person would also need to be in a group with Unrestricted data access or Sandboxed access to the relevant database (or table) to view that question.
- If the question was created using the native/SQL editor, the person would need to be a member of a group with both Unrestricted data access and Native query editing set to YES to view that question.
When you select Granular access for a database, you’ll be prompted to set permissions on the tables (or schemas) within that database. Here you’ll have some options, which differ depending on your Metabase plan.
Unrestricted access to the table
Groups with unrestricted access can use the graphical query builder to ask questions about this table.
No self-service access to the table
Groups with no self-service access to a table can’t access the table at all. They can, however, view questions that use data from that table, provided the group has access to the question’s collection, and they’re not in a group with blocked access for that table’s database.
Sandboxed access to the table
Sandboxed access to a table can restrict access to columns and rows of a table. Check out data sandboxing.
Native query editing
Members of a group with Native query editing set to “Yes” can:
This access level requires the group to additionally have Unrestricted data access for the database in question, since SQL queries can circumvent table-level permissions.
People in a group without Native query editing permissions will still be able to view the results of questions created from SQL/native queries (though just the results, not the query), or run an action, provided they 1) have collection access to the question or model, and 2) it doesn’t query a database that is blocked for that group.
You can set permissions on whether people in a group can download results (and how many rows) from a data source. Options are:
- No (they can’t download results)
- Granular (you want to set access for individual tables or schemas)
- 10 thousand rows
- 1 million rows
Manage table metadata
You can define whether a group can edit table metadata. Options are:
- Yes (meaning, they can edit metadata for that data source).
- Granular (to set permissions specific to each table).
The Manage database permission grants access to the settings page for a given database (i.e., the page at Admin settings > Databases > your database).
On the database settings page, you can:
Note that only admins can delete database connections in your Metabase, so people with Manage database permissions won’t see the Remove database button.