databricks Platform · Updated Jun 15, 2026

AUTO CDC for Streaming Tables

Streamline change data capture with automatic handling of out-of-order records.

What it does

The `FLOW AUTO CDC` clause allows users to create streaming tables in Databricks SQL that automatically process change data capture (CDC) records from a source. It simplifies the handling of out-of-order records and supports both slowly changing dimensions (SCD) type 1 and type 2 without the complexity of `MERGE INTO`. Users must set the `pipelines.channel` table property to 'PREVIEW' to utilize this feature.

Why it matters

This update significantly reduces the complexity and potential errors associated with processing CDC records, enhancing data accuracy and operational efficiency. It benefits data engineers and analysts by simplifying workflows, improving data freshness, and enabling more reliable historical tracking, which is crucial for analytics and reporting.

Key capabilities

Automatic Out-of-Order Handling

Automatically manages out-of-order records, reducing the need for complex logic.

Supports SCD Types

Allows for both SCD type 1 (direct updates) and SCD type 2 (history tracking) configurations.

Flexible CDC Processing

Enables users to specify keys, sequence columns, and conditions for delete and truncate operations.

Enhanced History Tracking

Provides options to track history on specific columns, improving data lineage and audit capabilities.

Prerequisites

Databricks Runtime 17.3 or above

Set the `pipelines.channel` table property to 'PREVIEW'

Recommended use cases

Real-time analytics on changing datasetsMaintaining historical records for compliance and reportingSimplifying ETL processes for CDC data

Limitations · remember

!Requires streaming sources; static sources may lead to errors.

!The `APPLY AS TRUNCATE WHEN` clause is only supported for SCD type 1.

Architecture

Data Source → Streaming Table → AUTO CDC Processing → Delta Lake

Get started

Set Table Property

Set the `pipelines.channel` property to 'PREVIEW' on the streaming table.

Create Streaming Table

Use the `CREATE OR REFRESH STREAMING TABLE` statement with the `FLOW AUTO CDC` clause.

Define Keys and Sequence

Specify the keys for identifying records and the sequence column for ordering.

Configure SCD Type

Choose between SCD type 1 or type 2 for storing records.

Implement History Tracking

Optionally track history on specific columns as needed.

Key takeaway

The AUTO CDC feature simplifies change data capture in Databricks SQL, enhancing data processing efficiency and accuracy.

2026-06-15-auto-cdc-flows-for-streaming-tables-in-databricks-sql-are-generally-available Generated by the Databricks Updates Agent · browse all updates