Roles and permissions
Minimum engine privileges required to run a Vulcan domain resource, engine by engine: PostgreSQL, Snowflake, Databricks, Spark, Trino, Redshift, and Athena.
Before a Vulcan domain resource can connect to a data engine, that engine must grant Vulcan's service account or principal a specific, minimal set of privileges: nothing more than what Vulcan actually does at each stage of a run.
This section is a reference for platform, database, and security administrators who need to grant those privileges correctly, engine by engine.
Supported engines
External Trino
SQL GRANT statements or connector access-control rules, on a client-managed Trino cluster
Dedicated Trino
No Trino-level grants (access control lives on each depot's underlying data source)
Amazon Athena
AWS IAM policies (and optionally Lake Formation) on the Vulcan service role, not SQL GRANTs
How the permission tiers work
The PostgreSQL, Snowflake, Databricks, Spark, and External Trino guides each break their privileges into three secret_name tiers (Dedicated Trino has no SQL grants of its own, and Redshift and Athena are covered in their engine guides), matched to what a Vulcan deployment actually needs:
<engine>-semantic-readonly
Vulcan only exposes existing tables or views through the semantic layer; no models run
<engine>-single-*-full
The source and target objects live inside a single database, catalog, or schema
<engine>-cross-*-full
Vulcan reads from one or more source catalogs and writes model outputs into a separate target
Grant the tier that matches the deployment topology, not more.
Where this fits
Consult the matching guide before pointing a Vulcan domain resource at a new data engine, or before rotating the credentials in a Vulcan gateway secret. Each guide documents exactly what Vulcan does: reads, writes, and internal state tracking, so the grants can be scoped to that behavior instead of granted broadly.
If a domain resource sets spec.trino:, DataOS provisions and connects to a Dedicated Trino cluster automatically. Read the Dedicated Trino guide first, since it changes where permissions apply. Otherwise, Vulcan connects directly to the engine named in the gateway configuration.
Next steps
PostgreSQL - Grant access for a PostgreSQL-backed deployment
Snowflake - Grant access for a Snowflake-backed deployment
Databricks - Grant access for a Databricks Unity Catalog deployment
Spark - Grant access for a DataOS Lakehouse deployment via Spark
External Trino - Grant access for a client-managed Trino cluster
Dedicated Trino - Understand the depot-level access model for a DataOS-managed Trino cluster
Last updated
Was this helpful?