Unity Catalog limitations and known issues

Summary

• Unity Catalog publishes its requirements and limitations in the official documentation, so the authoritative, always-current list lives in the Unity Catalog requirements and limitations docs.
• Compute requirements: clusters need Databricks Runtime 11.3 LTS or above and must use standard or dedicated access mode; SQL warehouses support Unity Catalog by default.
• Resource limits are quota-based, for example up to 32,768 columns per table, 10,000 tables per schema, and 1,000,000 tables per metastore; the full quota table is in the resource limits docs.
• Naming constraints: object names cannot contain periods, spaces, or forward slashes, and the name information_schema is reserved.
• Some features have runtime or access-mode conditions (for example bucketing is not supported, and certain UDF and view scenarios require specific Databricks Runtime versions).

Unity Catalog limitations and known issues

Unity Catalog is Databricks' unified governance layer for data and AI, and Databricks publishes its requirements and limitations directly in the product documentation. Because these details change as new Databricks Runtime versions and features ship, the authoritative and always-current lists are the Unity Catalog requirements and limitations page and the resource limits page. The summary below reflects what those docs describe; always confirm against the documentation for your Databricks Runtime version and cloud.

Requirements to use Unity Catalog

• Databricks Runtime: Unity Catalog is supported on clusters running Databricks Runtime 11.3 LTS or above. SQL warehouses support Unity Catalog by default on all versions.
• Access mode: Clusters must use standard or dedicated access mode to access Unity Catalog data. Unity Catalog is secure by default, so clusters without one of these access modes cannot access Unity Catalog data.
• Regions: Unity Catalog is available across AWS, Azure, and GCP regions.

Documented resource limits and quotas

Unity Catalog enforces quota-based limits at the metastore, catalog, schema, table, and model scopes. Examples from the resource limits documentation include:

• Up to 32,768 columns per table.
• Up to 10,000 tables per schema and 1,000,000 tables per metastore.
• Up to 10,000 schemas per catalog and 1,000 catalogs per metastore.
• Up to 10,000 volumes per schema and 100,000 volumes per metastore.
• Up to 10,000 functions per schema.
• Up to 1,000 registered models per schema and 5,000 per metastore, with up to 10,000 model versions per registered model.
• Up to 10,000 external locations per metastore and 1,000 storage and service credentials (combined) per metastore.
• Privilege limits of 4,000 on parent objects (metastores, catalogs, schemas) and 1,000 on non-parent objects (tables, views, shares).

These are examples; consult the docs for the complete quota table and any per-cloud differences.

Naming and identifier constraints

• Object names cannot contain periods (.), spaces, or forward slashes (/).
• The name information_schema is reserved and schemas with that name cannot be imported into a Unity Catalog metastore.

Feature-specific limitations

The requirements-and-limitations documentation lists conditions that depend on Databricks Runtime version and access mode. Documented examples include:

• Bucketing is not supported for Unity Catalog tables; commands that attempt to create bucketed tables raise an exception.
• Python UDFs (including UDAFs, UDTFs, and Pandas on Spark such as applyInPandas and mapInPandas) are not supported on Databricks Runtime 12.2 LTS and below; Python scalar UDFs are supported on Databricks Runtime 13.3 LTS and above.
• Scala UDFs are not supported on standard access mode compute on Databricks Runtime 14.1 and below; Scala scalar UDFs are supported on Databricks Runtime 14.2 and above.
• Scala thread pools require the special pools in org.apache.spark.util.ThreadUtils; standard Scala thread pools are not supported.
• R workloads do not support dynamic views for row-level or column-level security on Databricks Runtime 15.3 and below; use Databricks Runtime 15.4 LTS or above.
• Shallow clone is supported managed-table-to-managed-table on Databricks Runtime 13.3 LTS and above, and external-to-external on Databricks Runtime 14.2 and above; cross-type cloning between managed and external tables is not supported.
• Account-level groups are required for GRANT statements; groups created only at the workspace level cannot be used in Unity Catalog grants.
• External table partition operations (for example ALTER TABLE ... ADD PARTITION) require partition metadata logging to be enabled.

Where to find the authoritative, current list

Because Databricks adds capabilities and raises or removes limits over time, treat the documentation as the source of truth. Start with Unity Catalog requirements and limitations and resource limits, and check feature-specific pages (for example Delta Sharing has its own documented open sharing limitations). You can also learn more about the product on the Unity Catalog product page.

FAQs

What Databricks Runtime version do I need for Unity Catalog?

Databricks Runtime 11.3 LTS or above for clusters. SQL warehouses support Unity Catalog by default on all versions.

Are there limits on how many tables or catalogs I can create?

Yes. Documented quotas include up to 10,000 tables per schema, 1,000,000 tables per metastore, 10,000 schemas per catalog, and 1,000 catalogs per metastore. See the resource limits documentation for the full list.

Why can't I use certain characters in object names?

Unity Catalog object names cannot contain periods, spaces, or forward slashes, and information_schema is a reserved name. Choose names that avoid these characters.

Where is the definitive list of Unity Catalog limitations?

In the official documentation: the Unity Catalog requirements and limitations page and the resource limits page, plus feature-specific pages. These are kept current as the platform evolves.