Unable to Access Delta View from Azure Machine Learning via Delta Sharing – Is View Access Supported

[gg5]

Unable to Access Delta View from Azure Machine Learning via Delta Sharing – Is View Access Supported?

I am able to access the tables but while accessing the view I am getting below error.

Response from server:
{ 'details': [ { '@type': 'type.googleapis.com/google.rpc.ErrorInfo',
'domain': 'data-sharing.databricks.com',
'metadata': { 'dsError': 'DS_MATERIALIZATION_QUERY_FAILED',
'reason': 'Task '
'DATA-MATERIALIZATION-CREATE-EMPTY-TABLE '
'failed with message: An error '
'occurred during execution of task '
'Data materialization task: The '
'shared asset could not be '
'materialized due to asset not being '
'accessible in materialization '
'workspace. Please ask data provider '
'to contact Databricks support to '
'override the materialization '
'workspace. This caused all '
'downstream tasks to get skipped.',
'schema': 'xxxx',
'share': 'delta-share-xx-xx',
'table': 'xxxx'},
'reason': 'DS_MATERIALIZATION_QUERY_FAILED'},
{ '@type': 'type.googleapis.com/google.rpc.RequestInfo',
'request_id': 'xxxx-f8c2-4d85-9aa7-xxxx',
'serving_data': ''}],
'error_code': 'BAD_REQUEST',
'message': 'DS_MATERIALIZATION_QUERY_FAILED: Query failed for '
'xxx.xxxx from Share '
'delta-share-xx-xx.'}

Could someone suggest the steps to do enable the view access if it is possible?

[BigRoux]

Delta Sharing supports view sharing, but certain restrictions need to be considered. Views can be shared via Databricks-to-Databricks Delta Sharing, using Unity Catalog. Shared views must be defined on Delta tables or other shareable views and can only be queried under specific configurations. However, views referencing other shared tables or views are not supported.

The error "DS_MATERIALIZATION_QUERY_FAILED" indicates that the materialization of the shared asset (view) failed due to access or permissions issues in the materialization workspace. Cross-account view sharing requires serverless compute environments as views rely on materialization mechanisms to resolve their definitions.

Steps to Resolve the Issue: 1. Confirm Required Configurations: Ensure that the recipient environment is using a trusted compute such as a Serverless SQL warehouse. Views cannot benefit from shared-table materialization due to current technical limitations.

2. Override Materialization Workspace: Contact Databricks support to override the materialization workspace if the shared asset's materialization persists in failing.
3. Verify Permissions: Check that the system service principal has the required permissions to access referenced tables or views. If permission issues exist, work with the data provider to resolve them.
4. Follow Naming Guidelines: Make sure there are no catalog name conflicts between the provider-side and recipient-side namespaces.
5. Investigate Advanced Options: Consider dynamic fine-grained access controls or adjusting the filtering fleet implementation for shared views as an alternative approach.

Hope this helps, Lou.

[gg5]

Thanks for your quick response. Based on your explanation, I understand that accessing a Delta View from Azure Machine Learning via Delta Sharing is currently not supported due to the materialization limitations and trusted compute requirements. Unfortunately, in my business scenario, the need is to access a shared Delta View directly from AML — so just to confirm, this setup is not feasible at the moment, correct?

[BigRoux · Accepted solution]

Correct. Views are not supported (as of today) via Delta Sharing.  Cheers, Louis.