Databricks Asset Bundle (DAB) from a Git repo?

[dbx_687_3__1b3Q]

My earlier question was about creating a Databricks Asset Bundle (DAB) from an existing workspace. I was able to get that working but after further consideration and some experimenting, I need to alter my question. My question is now "how do I create a DAB from a Git repo?" 

We have a shared workspace where we've been building our project's Databricks assets (notebooks, python scripts, DLT pipelines, and workflows). The assets are eventually committed to our Azure DevOps Git repo. We create pull requests, code gets reviewed, and eventually the assets are merged into our release branch cleverly named "release". I need to use the release branch as the source for my DAB.

Based on the the example provided in the response, I dug a little deeper and discovered there is a workspace command to export an entire folder (export-dir). I modified the command to use export-dir and passed the path to the repo, "/Repos/release/repo_name_here". Along the way I also needed to change the arguments but, in the end, I was unable to successfully run the command because the specified path could not be found. I think it makes sense since it's likely looking in the workspace and I'm trying to use assets in a repo.

How do I build a DAB from a specific branch in an Azure DevOps Git repo?

[erima21]

Any help with this?

[erima21]

This is how I solved this:. Hope this works for you.
- bundle:

- jobs:

[Еmil]

Hi @erima21 

Do you run the job under service principal account? 
I have git authentication issue as it appears there is no way to get the git provider to use Microsoft Entra ID (formerly Azure Active Directory) authentication. When the job is executed it appears it can not access git...
I already posted about this issue here and I wonder if you had similar experience...

Thanks

[Еmil]

Sorry the link above doesn't work as my post on the databricks community forum was marked as spam. 
I have explained my issue on stackoverflow here.

[nicole_lu_PM]

> I have git authentication issue as it appears there is no way to get the git provider to use Microsoft Entra ID (formerly Azure Active Directory) authentication.

We published a joint-blog with MSFT to clarify the way to get Azure DevOps to use Microsoft Entra ID (formerly Azure Active Directory) authentication. Check this out!

https://www.databricks.com/blog/integrating-entra-id-azure-devops-and-databricks-better-security-cic...

[nicole_lu_PM · Accepted solution]

We are very close to having an end-to-end solution for deploying DABs from a Git folder (Repo) in the Workspace!

Check out my talk on DAIS24 here https://github.com/databricks/dais-cow-bff (video link on README). We are waiting for the feature that allows `databricks bundle deploy`  to work in the workspace web terminal. It should be available within the next month.

[AFDCO]

Hi, I'm in the same situation as you. Trying to export an entire workspace to deploy it in other environments. I'm having trouble especially with jobs, which I'm just able to do it with the api and not with bundle since it throws an error with paths --Error: Path (******) doesn't start with '/'--, which ironically is a requirement if you use git as source in the jobs.
It also export the pointed notebooks to the source paths, ignoring parent directories...

How did you manage to solve it?

Thank you 

[nicole_lu_PM]

Hi AFDCO,

> able to do it with the api and not with bundle since it throws an error with paths --Error: Path (******) doesn't start with '/'--, which ironically is a requirement if you use git as source in the jobs.

Bundles do deploy jobs that read from the Workspace, not Git.  For an example on the currently recommended approach, see this example: https://www.youtube.com/watch?v=DMwilNpDCiQ

[nicole_lu_PM]

`databricks bundle deploy` now works in the workspace web terminal without using a special version of Databricks CLI! 🔥