Deprecation of DBFS?

humayun
New Contributor II

‎02-24-2025 09:58 AM

I am using Azure Databricks DBFS. I have stored some python scripts, notebooks in folders in DBFS root and have mounted ADLS Gen2 to DBFS. Currently I do not have Unity Catalog enabled and this was setup 3-4 years ago. I came across this article 

https://docs.databricks.com/aws/en/dbfs/#:~:text=DBFS%20mounts%20are%20deprecated.,on%20the%20local%....

which mentions that Databricks does not recommend storing production data, libraries, or scripts in DBFS root and DBFS mounts are deprecated.

Is DBFS going to be deprecated altogether, the article specifically mentions about DBFS mounts. 

Do we have any specific date when these features will be completely removed from the platform.

Your suggestions, reference article and experiences will be helpful and I can plan things in advance.

Thanks.

Labels:

Advika_
Databricks Employee
Databricks Employee

‎02-27-2025 02:26 AM

Hello @humayun!

DBFS mounts and root storage are deprecated, as they are not secure with the Unity Catalog and pose security risks. While there’s no official end-of-life date, Databricks discourages using mounts and root storage. Existing setups will still work, but it's best to migrate to avoid future issues.

Kirankumarbs
Valued Contributor III

‎06-26-2026 05:51 AM

Hey @humayun 

Storing data DBFS provisioned storage is deprecated but not dbfs:/ URI scheme. Let's not get confused!

dbfs:/ scheme to access dbfs:/Volumes/${CATALOG}/libs/wheels is supported and not going to be deprecated!

Hope this helps and feel free to mark it as accepted solution so that other's can find it easily!!

0 Kudos

JeninaAngelin7
New Contributor III

‎06-28-2026 08:22 AM

In current state, when DBFS is disabled in a workspace, both the root filestore and mount paths become inaccessible.

What is deprecated are:

  • DBFS root storage the workspace‑provisioned filestore - /dbfs/FileStore/, /dbfs/root/ etc.

  • DBFS mounts (/dbfs/mnt/...) used to connect external storage. 

Best practices going forward:

  • Use external paths instead of mount paths - Azure: Use abfss://<container>@<storageaccount>.dfs.core.windows.net/... directly for ADLS Gen2. AWS: Use s3://<bucket>/<path> directly for S3.

  • Unity Catalog Volumes: Adopt Volumes for secure, governed access to external storage. These are accessed via dbfs:/Volumes/... and integrate with UC permissions.

  • System paths: Internal locations like MLflow artifact storage (dbfs:/databricks/mlflow/...) remain supported and are managed by Databricks services.

0 Kudos