October 2021
These features and Databricks platform improvements were released in October 2021.
Note
Releases are staged. Your Databricks account may not be updated until a week or more after the initial release date.
Databricks Runtime 8.2 series support ends
October 22, 2021
Support for Databricks Runtime 8.2 and Databricks Runtime 8.2 for Machine Learning ended on October 22. See Databricks support lifecycles.
Databricks Runtime 10.0 and 10.0 ML are GA; 10.0 Photon is Public Preview
October 20, 2021
Databricks Runtime 10.0 and 10.0 ML are now generally available. 10.0 Photon is in Public Preview.
See the release notes at Databricks Runtime 10.0 (EoS) and Databricks Runtime 10.0 for ML (EoS).
Limit the set of VPC endpoints your workspace can use for AWS PrivateLink connections (Public Preview)
October 19, 2021
For AWS PrivateLink connections, the private_access_level
configuration field lets you narrow the set of VPC endpoints that work with each workspace. The default behavior is not to filter the VPC endpoints that can connect to your workspace using AWS PrivateLink, which is unchanged. The new configuration allows you to specify which AWS PrivateLink VPC endpoints can connect to the workspace. You can choose from two new connection levels: limit connections to those VPC endpoints that are registered in your Databricks account, or limit connections to an explicit set of VPC endpoints.
There is also a new operation in the Account API to update an existing private access settings object. Use this to modify PrivateLink settings for an existing workspace that already has an associated private access settings object.
User interface improvements for Delta Live Tables (Public Preview)
October 18-25, 2021: Version 3.57
This release includes the following enhancements to the Delta Live Tables UI:
A Notebook not found message is now displayed in the Create Pipeline dialog or the datasets detail panel when an invalid notebook path is provided.
You can now provide feedback on Delta Live Tables by clicking the Provide Feedback link on the Pipelines page or the Pipeline Details page. When you click the Provide Feedback link, a customer feedback survey opens in a new window.
Specify a fixed-size cluster when you create a new pipeline in Delta Live Tables (Public Preview)
October 18-25, 2021: Version 3.57
You can now create a cluster for your Delta Live Tables pipeline with a fixed number of worker nodes, providing more control over the cluster resources used by the pipeline. To create a cluster with a fixed number of nodes, disable Enable autoscaling and enter the number of nodes in the Workers field when you create a new pipeline.
View data quality metrics for tables in Delta Live Tables triggered pipelines (Public Preview)
October 18-25, 2021: Version 3.57
You can now see data quality metrics for tables when your pipeline runs in triggered mode, including the number of records written, the number of records dropped, and the number of records that passed or failed each data quality constraint. To view data quality metrics, select the table in the Graph tab on the Pipeline Details page.
Jobs orchestration is now GA
October 14, 2021
Databricks is pleased to announce the general availability of Databricks jobs orchestration. Jobs orchestration allows you to define and run a job with multiple tasks, simplifying the creation, scheduling, execution, and monitoring of complex data and machine learning applications. Jobs orchestration needs to be enabled by an administrator and is disabled by default. See Schedule and orchestrate workflows.
Databricks is also pleased to announce general availability of version 2.1 of the Jobs API. This version includes updates that fully support the orchestration of multiple tasks with Databricks jobs. See Updating from Jobs API 2.0 to 2.1 for information on updating clients to support jobs that orchestrate multiple tasks.
Databricks Connector for Power BI
October 13, 2021
A new version of the Power BI connector is now available. This release adds support for navigation through three-level namespaces in the Unity Catalog, ensures that query execution can be cancelled, and enables native query passthrough for reduced latency on Databricks SQL and Databricks Runtime 8.3 and above.
Repos now supports arbitrary file types
October 7, 2021
In addition to syncing notebooks with a remote Git repository, Files in Repos lets you sync any type of file, such as .py
files, data files in .csv
or .json
format, or .yaml
configuration files. You can import and read these files within a Databricks Repo. You can also view and edit text-encoded files in the UI. For more information, see Git integration for Databricks Git folders.
More detailed job run output with the Jobs API
October 4-11, 2021: Version 3.56
The response from the Jobs API Runs get output request is updated with new fields to return additional output, logging, and error detail for job runs.
Improved readability of notebook paths in the Jobs UI
October 4-11, 2021: Version 3.56
Notebook paths are now truncated from the left when viewing job details in the jobs UI, ensuring visibility of the notebook name. Previously, notebook paths were truncated from the right, often obscuring the notebook name.
Open your Delta Live Tables pipeline in a new tab or window
October 4-11, 2021: Version 3.56
Pipeline names are now rendered as a link when you view the pipelines list in the Delta Live Tables UI, providing access to context menu options such as opening the pipeline details in a new tab or window.
New escape sequence for $
in legacy input widgets in SQL
October 4-11, 2021: Version 3.56
To escape the $
character in legacy input widgets in SQL, use \$
. If you have used $\
in existing widgets, it continues to work, but Databricks recommends that you update widgets to use the new escape sequence. See Databricks widgets.
Faster model deployment with automatically generated batch inference notebook
October 4-11, 2021: Version 3.56
After a model is registered in Model Registry, you can automatically generate a notebook to use the model for batch inference or create an endpoint to use the model for real-time serving. For details, see Use model for inference.