Databricks Syncs Postgres to Lakehouse Natively

Databricks is rolling out Native Lakehouse Sync, a feature designed to bypass traditional data pipelines for moving operational data into its Lakehouse platform. The company announced the public preview of this capability, which replicates data from Lakebase Postgres directly into Unity Catalog managed tables. This new approach aims to simplify data integration for modern AI and analytics applications. According to the announcement, the sync is a native property of Lakebase, eliminating the need for external compute or complex pipeline configurations.

Historically, moving data from operational databases to data warehouses or analytics platforms involved intricate Change Data Capture (CDC) stacks or batch processes. Databricks argues these methods falter with the rise of agent-first development, which relies on rapid data branching and scaling to zero. Traditional 'zero-ETL' solutions often assume stable workloads and predictable query volumes, assumptions that break down in dynamic agent-driven environments.

Why a Native Approach?

The core of Databricks' new offering hinges on Lakebase running on the same open, low-cost cloud storage as the Lakehouse. This shared storage foundation allows data movement to become an intrinsic database function, rather than an external process.

Native Lakehouse Sync decodes Lakebase's Write-Ahead-Log (WAL) and writes directly to Unity Catalog Managed Tables. Enabling this sync is reportedly a simple, schema-level toggle that takes less than a minute.

Databricks claims this process has zero impact on Postgres performance and incurs no additional cost. Because Databricks controls both the source (Lakebase) and destination (Unity Catalog), schema changes are propagated automatically, mitigating common issues of data drift and lag.

Agent-First Benefits

For agent-first development, Native Lakehouse Sync inherits key behaviors. It scales down when the database scales to zero and resumes from the last recorded state. All monitoring and observability are managed within the Lakebase project itself.

Schema propagation is automatic. Adding a column to a source table instantly reflects in the destination, and dropping a column retains it in the destination, preventing agents from needing to reconfigure syncs.

This is a significant shift from previous approaches, where Databricks Lakehouse Gets Postgres Boost on Azure required more intricate setup.

Lakehouse Capabilities at the Destination

Once data lands in Unity Catalog managed tables, it immediately gains access to the full suite of Lakehouse capabilities.

This includes AI-native analytics, making data queryable by agents like Databricks Genie and Genie Code. The data is universally readable by Spark, Databricks SQL, and other tools supporting Delta or Iceberg formats.

Unified governance features like lineage, access policies, and audits are inherited from Unity Catalog. Databricks' optimization features, such as Predictive Optimization and Liquid Clustering, apply automatically.

Crucially, every insert, update, and delete is captured as SCD Type 2 history by default, providing built-in versioning, audit logs, and Change Data Capture (CDF) semantics without extra configuration. This contrasts with older methods where Lakebase Postgres sync could require more manual tuning.

Unlocking New Use Cases

The combination of these features enables previously complex patterns:

• Agentic Memory and Live ML Features: Application writes are available in Unity Catalog within minutes, allowing models to retrain and score against the most current application state.
• Operational Data in Medallion Architecture: Lakebase can serve as the Bronze Tables layer, with high-velocity updates captured and their full history automatically flowing into the Lakehouse as SCD Type 2.
• Compliance and Audit: Every data modification is logged as historical data in Unity Catalog, eliminating the need for separate application-side history tracking or audit pipelines.

Native Lakehouse Sync is now in public preview. Databricks emphasizes that setting up a Lakebase is instant, and toggling sync on a schema makes all existing and future tables appear in Unity Catalog within a minute.