Multi-tenancy on Superset

[sijonelis]

Hi all, I am trying to think of a way to implement superset in a multi-tenant scenario where each tenant has their own schema (data source).

The tables are identical for each tenant, can can be reliably identified, such as having a suffix at the end of the table name, i.e. demo_table_t_001, demo_table_t_002 and so on.

From the initial investigation, it seems that superset only supports RLS on a single dashboard/slice which is not viable for the tenant isolation level that i am running on.

Keeping this in mind i have come up with the following poc architecture, introducing an intermediate layer between the superset and the product's frontend and adding additional logic to maintain a set of dashboards for each tenant, derived from and synced to the "master" dashboard.

While it works on paper, it seems to me that this is a very complicated approach to solve such an issue that can potentially introduce scaling issues down the line. At the same time seems to be the best case approach as Superset's slice is tightly coupled with the data source, making any tinkering with the superset code an even more complicated approach.

Has anyone in the community met such a usecase and have managed to apply Superset to solve it?