Access mode for pool compute

[IliaSinev]

Is there a way to set Access Mode: Shared to pool instances similar to All Purpose or Job clusters?

We are getting an error reading from a table with a masking set up on a column:

Failed to acquire a SAS token for list on /schema1/table1/_delta_log due to com.databricks.sql.managedcatalog.UnityCatalogServiceException: [RequestId=7d2e1b1b-005d-4110-8b56-4c8e9f92ce9d ErrorClass=INVALID_PARAMETER_VALUE.ROW_COLUMN_ACCESS_POLICIES_NOT_SUPPORTED_ON_ASSIGNED_CLUSTERS] Query on table catalog1.schema1.table1 with row filter or column mask not supported on assigned clusters.

 Similar errors were already discussed for All Purpose and Job clusters, and the working solution is to change the access mode to Shared. I don't see this setting however for pool instances - is there a way to work around?

[Brahmareddy]

Hi IliaSinev,

How are you doing today?, As per my understanding, You're right—Databricks doesn’t currently let you set Shared access mode on clusters that use instance pools, which is why you’re seeing that error with column masking and row-level filters. These security features only work when the cluster is in Shared mode, because it needs to know which user is running the query. Instance pools always use Assigned mode, so they’re not compatible with these features. The best workaround for now is to use a Shared all-purpose or job cluster (without a pool) just for the workloads that need masking or row filters. It’s not as fast to start, but it will work. Let me know if you want help setting that up!

Regards,

Brahma

[IliaSinev]

Hi @Brahmareddy, thanks for reply. It seems that a higher Runtime version could help: https://learn.microsoft.com/en-us/azure/databricks/compute/access-mode-limitations#fine-grained-acce... I am going to try it on Monday and report back