Lakebase Change Data Feed

Lakebase Change Data Feed is the current Databricks name, previously Lakehouse sync (since May 2026).

Lakehouse sync → Lakebase Change Data Feed

Captures every insert, update, and delete on a Lakebase Postgres table from the write-ahead log and stores each as a new row in a Unity Catalog managed Delta table (one lb_<table>_history table per source), flushed about every 15 seconds, for downstream pipelines and audit.

• It runs on a Postgres extension called wal2delta, a name that doubles as the entire job description - read the WAL, write Delta. 🔗
• Do not turn on Delta Lake's own change data feed on its destination tables - a change feed of the change feed breaks the re-snapshot it runs when the source schema changes. 🔗
• An empty table stays out of the feed entirely until it holds at least one row, on the reasonable grounds that nothing has changed yet. 🔗

⚠ Limitations: Source tables need REPLICA IDENTITY FULL, partitioned tables are not supported, empty tables are skipped until they hold a row, destination history tables must not get row filters, column masks, or Delta change data feed, and it does not work when the destination catalog's managed storage is reachable only through a private endpoint. 🔗

Category

Data engineering

In use from

May 2026🔗

Also known as

Lakebase CDF

Verified

2026-10-14