Database Branching in Postgres: Git-Style Workflows with Databricks Lakebase

This post introduces database branching in Databricks Lakebase Postgres as a copy-on-write primitive that enables Git-like isolated database environments for development workflows.

• •Traditional pg_dump-based cloning scales linearly with database size (minutes to hours), while Lakebase branches are created in seconds regardless of database size
• •Lakebase separates compute from storage using a log-structured, versioned storage engine; branches share underlying data and only write changes separately (copy-on-write)
• •Each branch gets its own Postgres connection string and independent compute endpoint, with idle branches auto-scaling to zero
• •Key workflow patterns include one branch per developer, one per pull request (integrated with Vercel/Netlify preview deployments), one per CI test run, and ephemeral environments for AI agents