Hi @Suheb,
That’s a great question, version control is one of the most important things to get right early on.
As a best practice, you should never run notebooks directly in production. Instead, notebooks should be treated as development assets, once validated, they should be packaged, version-controlled, and deployed through proper CI/CD.
1. Use Git integration
Databricks integrates directly with GitHub, Azure DevOps and GitLab.
Always link your workspace to a Git repo and commit your notebook changes regularly, this keeps full version history and supports collaboration.
2. Package and deploy, don’t run manually
Convert notebooks into production-ready code (Python modules or .py scripts).
Use Databricks Asset Bundles (DAB) or your CI/CD pipeline to deploy jobs, pipelines, and workflows, not raw notebooks.
This ensures environments (Dev, Test, Prod) stay consistent and auditable.
3. Automate with Workflows
Use Jobs or Workflows to orchestrate your pipelines instead of manual runs.
Parameters, retries, and alerts can all be managed centrally.
4. Keep documentation handy
Databricks provides extensive documentation for both Git integration and CI/CD with DAB, plenty of examples depending on your setup (GitHub Actions, Azure DevOps, Jenkins, etc.).
In short:
Develop in notebooks, version in Git, deploy with DAB, and run in production via Jobs/Workflows never directly from a notebook.
