GitHub version control

Note

To sync your work in Databricks with a remote Git repository, Databricks recommends using Repos for Git integration.

This article describes how to set up version control for notebooks using GitHub through the UI. You can also use the Databricks CLI or Workspace API to import and export notebooks and manage notebook versions using GitHub tools.

Enable and disable Git versioning

By default version control is enabled. To toggle this setting, see Manage the ability to version notebooks in Git. If Git versioning is disabled, the Git Integration tab is not available in the User Settings screen.

Configure version control

To configure version control, you create access credentials in your version control provider, then add those credentials to Databricks.

Get an access token

In GitHub, follow these steps to create a personal access token that allows access to your repositories:

  1. In the upper-right corner of any page, click your profile photo, then click Settings.

  2. Click Developer settings.

  3. Click the Personal access tokens tab.

  4. Click the Generate new token button.

  5. Enter a token description.

  6. Select the repo permission, and click the Generate token button.

    Generate GitHub token
  7. Copy the token to your clipboard. You enter this token in Databricks in the next step.

See the GitHub documentation to learn more about how to create personal access tokens.

Save your access token to Databricks

  1. In Databricks, click the User icon Account Icon at the top right of your screen and click User Settings.

    AccountSettings
  2. Click the Git Integration tab.

  3. If you have previously entered credentials, click the Change token or app password button.

  4. In the Git provider drop-down, select GitHub.

    Select GitHub Git provider
  5. Paste your token into the Token or app password field and click Save.

Work with notebook revisions

You work with notebook revisions in the history panel. Open the history panel by clicking Revision history at the top right of the notebook.

Revision history

Note

You cannot modify a notebook while the history panel is open.

Save a notebook to GitHub

While the changes that you make to your notebook are saved automatically to the Databricks revision history, changes do not automatically persist to GitHub.

  1. Click Revision history at the top right of the notebook to open the history Panel.

    History panel - save now
  2. Click Save Now to save your notebook to GitHub. The Save Notebook Revision dialog displays.

  3. Optionally, enter a message to describe your change.

  4. Make sure that Also commit to Git is selected.

    Save revision
  5. Click Save.

Revert or update a notebook to a version from GitHub

Once you link a notebook, Databricks syncs your history with Git every time you re-open the history panel. Versions that sync to Git have commit hashes as part of the entry.

  1. Click Revision history at the top right of the notebook to open the history Panel.

    History panel
  2. Choose an entry in the history panel. Databricks displays that version.

  3. Click Restore this version.

  4. Click Confirm to confirm that you want to restore that version.

Branch support

You can work on any branch of your repository and create new branches inside Databricks.

Create a branch

  1. Click Revision history at the top right of the notebook to open the history Panel.

  2. Click the Git status bar to open the GitHub panel.

  3. Click the Branch dropdown.

  4. Enter a branch name.

    Create branch
  5. Select the Create Branch option at the bottom of the dropdown. The parent branch is indicated. You always branch from your current selected branch.

Create a pull request

  1. Click Revision history at the top right of the notebook to open the history Panel.

  2. Click the Git status bar to open the GitHub panel.

    Git preferences - create pull request
  3. Click Create PR. GitHub opens to a pull request page for the branch.

Rebase a branch

You can also rebase your branch inside Databricks. The Rebase link displays if new commits are available in the parent branch. Only rebasing on top of the default branch of the parent repository is supported.

Rebase

For example, assume that you are working on databricks/reference-apps. You fork it into your own account (for example, brkyvz) and start working on a branch called my-branch. If a new update is pushed to databricks:master, then the Rebase button displays, and you will be able to pull the changes into your branch brkyvz:my-branch.

Rebasing works a little differently in Databricks. Assume the following branch structure:

Before rebase branch structure

After a rebase, the branch structure will look like:

After rebase branch structure

What’s different here is that Commits C5 and C6 will not apply on top of C4. They will appear as local changes in your notebook. Any merge conflict will show up as follows:

Merge conflict

You can then commit to GitHub once again using the Save Now button.

What happens if someone branched off from my branch that I just rebased?

If your branch (for example, branch-a) was the base for another branch (branch-b), and you rebase, you need not worry! Once a user also rebases branch-b, everything will work out. The best practice in this situation is to use separate branches for separate notebooks.

Best practices for code reviews

Databricks supports Git branching.

  • You can link a notebook to any branch in a repository. Databricks recommends using a separate branch for each notebook.
  • During development, you can link a notebook to a fork of a repository or to a non-default branch in the main repository. To integrate your changes upstream, you can use the Create PR link in the Git Preferences dialog in Databricks to create a GitHub pull request. The Create PR link displays only if you’re not working on the default branch of the parent repository.

GitHub Enterprise

Preview

This feature is in Private Preview. To try it, reach out to your Databricks contact.

You can also use the Workspace API to programmatically create notebooks and manage the code base in GitHub Enterprise Server.

Troubleshooting

If you receive errors related to syncing GitHub history, verify the following:

  • You can only link a notebook to an initialized Git repository that isn’t empty. Test the URL in a web browser.
  • The GitHub personal access token must be active.
  • To use a private GitHub repository, you must have permission to read the repository.
  • If a notebook is linked to a GitHub branch that is renamed, the change is not automaticaly reflected in Databricks. You must re-link the notebook to the branch manually.