Skip to main content
DatabricksPySparkSnowflakeApache Airflow

Databricks Catalog, Schema & Table Permissions (RBAC)

🔐 A Simple Story — “Too Many People, Too Much Data”​

Imagine a company where:

  • Analysts need access to Gold tables
  • Data engineers need access to Bronze/Silver
  • Data scientists need feature tables
  • Finance wants restricted datasets
  • Interns should not see sensitive customer info

Without proper access controls, your Databricks environment becomes:

⚠️ Messy
⚠️ Risky
⚠️ Hard to audit

What you need is a clean, structured permission system that controls who can access what — and keeps everything consistent.

This is where RBAC (Role-Based Access Control) in Databricks shines.

🗂 Understanding the Structure: Catalog → Schema → Table​

Databricks organizes storage and permissions in three layers:


CATALOG
└── SCHEMA
└── TABLE / VIEW / FUNCTION

Think of it like:

  • Catalog = entire building
  • Schema = floor inside the building
  • Table = rooms inside the floor

You can set permissions at any level.

🏢 1. Catalog — The Top Level​

A catalog contains schemas and tables.
In Unity Catalog, this is the highest namespace.

Example:


hive_metastore
main
sales_catalog
ml_catalog

Permissions Often Set Here:​

  • USE CATALOG
  • CREATE SCHEMA
  • OWN
  • READ FILES (for external locations)

Example:​

GRANT USE CATALOG ON CATALOG sales_catalog TO analyst_team;

This allows the team to see and navigate the catalog.

🗄 2. Schema — Organizing Tables​

A schema groups related objects.

Examples:

sales_catalog
    ├── bronze
    ├── silver
    └── gold

Permissions Often Set Here:​

  • USE SCHEMA
  • CREATE TABLE
  • SELECT
  • MODIFY
  • OWN

Example:​

GRANT USE SCHEMA ON SCHEMA sales_catalog.gold TO bi_team;

Now BI users can query tables inside the Gold schema.

📊 3. Table — The Most Detailed Level​

Tables hold actual data.

Examples:

sales_catalog.gold.daily_revenue
sales_catalog.silver.cleaned_orders

Permissions Often Set Here:​

  • SELECT (read data)
  • MODIFY (insert/update/delete)
  • OWN (full control)

Example:​

GRANT SELECT ON TABLE sales_catalog.gold.daily_revenue TO finance_team;

Now finance users can read the Gold revenue table — and only that table.

🧠 Important RBAC Concepts​

✔ Ownership​

Every object (catalog, schema, table) has an owner.

Owners can:

  • Modify
  • Grant permissions
  • Drop objects

You typically assign ownership to:

  • Admin groups
  • Data engineering teams

✔ Principle of Least Privilege​

Always give only the access required:

  • Analysts → Gold
  • Data Scientists → Feature tables
  • Engineers → Bronze/Silver
  • Finance → Restricted Gold tables
  • Interns → Only learning datasets

This keeps your environment safe and compliant.

✔ Inheritance​

If you grant a privilege at a higher level, it flows downward.

Example:

GRANT SELECT ON SCHEMA sales_catalog.gold TO bi_team;

This gives read access to all tables inside gold.

🧪 Practical Real-World Permission Patterns​

Pattern 1 — BI/Analytics Teams​

Only Gold data is needed.

GRANT SELECT ON SCHEMA sales_catalog.gold TO bi_team;

Pattern 2 — Data Engineers​

Full control of Bronze/Silver.

GRANT MODIFY ON SCHEMA sales_catalog.bronze TO data_eng;
GRANT MODIFY ON SCHEMA sales_catalog.silver TO data_eng;

Pattern 3 — Secure Financial Data​

Only the finance group can read.

GRANT SELECT ON TABLE sales_catalog.gold.finance_metrics TO finance_team;

Pattern 4 — External Tables​

Need READ FILES privilege.

GRANT READ FILES ON EXTERNAL LOCATION finance_location TO finance_team;

🛡 Unity Catalog Makes RBAC Easy​

Unity Catalog centralizes permissions across:

  • Databricks SQL
  • Notebooks
  • Jobs
  • MLflow
  • Delta Lake

So the same RBAC rules apply everywhere — clean, consistent, reliable.

📘 Summary​

  • Databricks uses Catalog → Schema → Table for organizing data.
  • RBAC (Role-Based Access Control) lets you define who can do what.
  • You can control access at any level: catalog, schema, or table.
  • Permissions include: USE, SELECT, MODIFY, CREATE, OWN, etc.
  • Unity Catalog provides centralized, secure, enterprise-grade permission management.
  • Always follow the least privilege model for safety and compliance.

RBAC ensures the Lakehouse stays clean, secure, and easy to manage as your organization grows.

👉 Next Topic

Databricks LakeFlow — Unified ETL, Orchestration & Governance