access to event_log TVF

[giladba]

Hi, 

According to the documentation:

https://docs.databricks.com/en/delta-live-tables/observability.html

"The event_log TVF can be called only by the pipeline owner and a view created over the event_log TVF can be queried only by the pipeline owner. The view cannot be shared with other users."

Can you suggest a workaround?

After searching there is another documentation that says"
https://learn.microsoft.com/en-us/azure/databricks/sql/language-manual/functions/event_log
"Only owners of the pipeline, streaming table, or materialized view can view the event log. Create a view and grant users access on the view to allow other users to query the event log."

> CREATE VIEW event_log_raw AS SELECT * FROM event_log(table(my_mv));
> GRANT SELECT ON VIEW event_log_raw TO `user@databricks.com`;

After granting users access on the view I created,  the user still can't access the view I created.

Would appreciate your help!

Thanks

[Atanu]

@giladba What is the error the user is getting after granting the permission?

[giladba]

com.databricks.backend.common.rpc.SparkDriverExceptions$SQLExecutionException: com.databricks.sql.managedcatalog.acl.UnauthorizedAccessException: PERMISSION_DENIED: User does not have permission to access event logs of pipeline 'pipelineid'

[Kaniz_Fatma]

Hi @giladba, The error message "com.databricks.backend.common.rpc.SparkDriverExceptions$SQLExecutionException: com.databricks.sql.managedcatalog.acl.UnauthorizedAccessException: PERMISSION_DENIED: User does not have permission to access event logs of pipeline '_pipelineid_'"  indicates that the user does not have the necessary permissions to access the event logs of the specified pipeline. This could be due to the user not having the correct permissions for the securables in the metastore, which can include catalogs, schemas, tables, views, etc.
• The user not being registered in the account console to access the Unity Catalog.
• The user does not have the "SELECT" permission on certain tables.

To resolve this issue, you can validate if the user/service principal has the proper permission to access the event logs of the pipeline.
• Verify if the user is registered in the account console. If not, register the user.
• Check if the user has "SELECT" permission on the necessary tables. If not, grant the required permissions using the GRANT SELECT ON <table_name> TO <username> command.

[giladba]

Thanks. The user is part of a group that has select on the relevant view and all the parent permissions needed for the Unity Catalog.

[neha_ayodhya]

Hi,
I am also facing the same issue, even after following all the steps mentioned, I am not able to query the event logs.
any help will be greatly appreciated.

[hcjp]

Hello,

@Kaniz_Fatma We are using Databricks on Azure but I suspect this issue impacts all cloud providers.  Essentially, the event_log Table Value Function is only allowing Pipeline Owner to view logs and this is too restrictive.  Despite documentation that states the owner can grant SELECT privileges on a view to other users/groups, this functionality does not work (error is unauthorized as documented above): https://learn.microsoft.com/en-us/azure/databricks/sql/language-manual/functions/event_log#usage

The documented approach at the link above likely doesn't work as users would need to also be granted SELECT on the underlying object (Table?).  In our scenario, we have a Service Principal running our DLT pipelines and need our Data Engineers (group) to be able to view (troubleshoot, monitor, etc.) the event_log of these pipelines.

Would you please raise this as a bug (if intended functionality is to allow granting select permissions) or as a new feature as the current capability is too restrictive.

Resolving this issue will unlock a lot of value for the Databricks community as the event_log contains highly valuable information.

Thank you.

[larsbbb]

Is there already a solution for this?

[hcjp]

As per this documentation, https://learn.microsoft.com/en-us/azure/databricks/delta-live-tables/unity-catalog, the issue here is documented as a current Limitation:

Limitations

The following are limitations when using Unity Catalog with Delta Live Tables:

• You cannot use the event_log table valued function in a pipeline or query to access the event logs of multiple pipelines. (hcjp: not directly related to this thread but also delivers value when limitation removed)

• You cannot share a view created over the event_log table valued function with other users. (hcjp: limitation associated with this specific issue)

@Kaniz_Fatma - I think the community ask, or at least those on this thread, is to get some visibility into when these limitations will be removed.  Is this currently on the backlog and, if not, is there some way we can upvote it?

Thank you.

[larsbbb]

Thank you for summarizing the issues @hcjp! Looking forward to the reply of @Kaniz_Fatma 

[Kaniz_Fatma]

Hi @larsbbb and @hcjp, You can submit feedback directly to the product team to influence the Databricks product roadmap in the following ways:

• To quickly submit feedback about your experience with Databricks, fill out the feedback form in your workspace.

• To interactively contribute to the product roadmap, submit a feature request in the Ideas Portal. You can view, comment, and vote up other users’ requests. You can also monitor the progress of your favorite ideas as the Databricks product team goes through their product planning and development process.

To work around this limitation, a few suggestions are provided:

1. Create a notebook or SQL script that queries the event log and contains the necessary logic to extract the desired information. The pipeline owner can then share this notebook or script with other users, allowing them to run it and access the event log data indirectly.
2. Explore the possibility of creating a separate table or dataset that contains the relevant event log data, extracted and transformed as needed. The pipeline owner can then grant SELECT access on this table to the desired users, enabling them to directly query the data without accessing the event log directly.