databricks Platform · Updated Jun 12, 2026

Liquid Clustering Conversion

Convert partitioned Delta Lake tables to liquid clustering with minimal downtime.

What it does

This feature allows users to convert existing partitioned Delta Lake tables to liquid clustering using the `ALTER TABLE ... REPLACE PARTITIONED BY WITH CLUSTER BY` command. It minimizes reader and writer downtime during the conversion process and supports both external and managed tables. Liquid clustering optimizes query performance by automatically organizing data based on clustering keys, making it easier to adapt to changing analytic needs.

Why it matters

This update enhances data management efficiency and query performance, particularly for organizations dealing with high cardinality columns and heavy data skew. By simplifying the data layout and reducing maintenance efforts, businesses can achieve better performance and lower operational costs, ultimately improving productivity and responsiveness to changing data access patterns.

Key capabilities

Minimized Downtime

Conversion minimizes reader and writer downtime during the transition.

Flexible Clustering Keys

Allows redefining clustering keys without rewriting existing data.

Performance Optimization

Improves query performance by automatically organizing data based on clustering keys.

Support for Streaming

Liquid clustering applies to both streaming tables and materialized views.

Prerequisites

Databricks Runtime 18.1 or above for conversion.

Existing Delta Lake table must be partitioned.

Recommended use cases

Optimizing queries on high cardinality columns.Managing tables with heavy data skew.Supporting fast-growing tables with concurrent write requirements.Adapting to varied or changing access patterns.

Limitations · remember

!Streaming tables and materialized views in Lakeflow Spark Declarative Pipelines are not supported.

!Tables using Delta Sharing with partition filtering are not supported.

Architecture

Delta Lake Table → Liquid Clustering → Query Optimization → Data Management

Get started

Run Conversion Command

Use `ALTER TABLE REPLACE PARTITIONED BY WITH CLUSTER BY ()`.

Optimize Table

Run `OPTIMIZE ` to apply clustering after conversion.

Verify Conversion

Run `DESCRIBE EXTENDED ` to check new clustering columns.

Handle Concurrent Workloads

Ensure Databricks Runtime 15.4 LTS or above for concurrent writes during conversion.

Rollback if Necessary

Use `RESTORE` to revert to the previous version if needed.

Key takeaway

The ability to convert partitioned tables to liquid clustering enhances performance and simplifies data management.

2026-06-12-converting-a-partitioned-table-to-liquid-clustering-is-generally-available Generated by the Databricks Updates Agent · browse all updates