Promoting one codebase across dev, test, and prod with bundle variables and overrides

Databricks Asset Bundles let you define one codebase and deploy it consistently across dev, test, and prod using a single databricks.yml with target-specific overrides. Variables parameterize values like cluster sizes, paths, and job schedules so each target gets appropriate settings without duplicating code. This keeps environment differences declarative and version-controlled rather than hardcoded.

1 · Learn the must-know

• 1Bundle configuration is defined in databricks.yml, with a top-level variables block declaring defaults and a targets block (e.g., dev, staging, prod) that can override those variables or resource settings per environment.
• 2The mode: development setting on a target enables features like resource name prefixing and auto-pausing schedules to prevent dev runs from colliding with production resources.
• 3Variables can be overridden via the CLI at deploy time using --var, via environment variables, or in the target-specific variables block, with CLI/env taking precedence over file defaults.
• 4Each target typically maps to a distinct workspace (via host) and run_as identity, so the same job/pipeline definitions deploy to physically separate dev/test/prod workspaces or paths.
• 5Complex variables (maps/objects, e.g., cluster spec overrides) can be overridden per target, not just simple strings, allowing structural differences like node types or autoscaling between environments.
• 6Running databricks bundle validate before deploy catches misconfigured variable references or target overrides early, and databricks bundle deploy -t <target> deploys to the specified environment explicitly.

2 · Check your understanding