Azure Databricks Serverless queries hang or fail with PERMISSION_DENIED: Cert validation failed. Origin workspace. Even SELECT 1 fails, while basic Python like print("hello") works.

Hi @Yogesh Niture ,
Based on the troubleshooting results, this appears to be isolated to the Azure Databricks Serverless notebook compute path, rather than a Unity Catalog permission or table-level issue.

Since:

• Python execution works normally.
• SELECT 1 and USE CATALOG fail/hang only on Serverless.
• The same Unity Catalog operations work successfully from a SQL Warehouse.
• A fresh notebook and Serverless session produce the same behavior.
• The error is PERMISSION_DENIED: Cert validation failed. Origin workspace.

I would recommend checking the workspace's Serverless networking/security configuration, particularly any serverless network policies, firewall/NSP rules, private connectivity, or other network controls that could affect Serverless compute. Azure Databricks Serverless uses a managed compute plane, and its connectivity to Azure resources can be controlled through Network Connectivity Configurations (NCCs) and serverless network policies.

However, because SELECT 1 itself fails and the SQL Warehouse works, if no customer-side network restriction is identified, this should be escalated to Microsoft/Azure Databricks Support as a possible Serverless backend/control-plane certificate validation issue.

When opening the support request, provide:

1. Workspace URL/ID and Azure region
2. Exact UTC timestamp of the failure
3. The full PERMISSION_DENIED: Cert validation failed. Origin workspace error
4. Confirmation that SQL Warehouse can access the same catalog successfully
5. Confirmation that a new Serverless session and notebook were tested
6. Screenshot/logs showing the failure

Ask support to specifically investigate the Serverless notebook compute → workspace/Unity Catalog service certificate validation path.

This distinction is important because Azure Databricks documentation notes that SSL/certificate failures can be caused by networking/security configuration, while Serverless compute has its own managed networking pathBased on the troubleshooting results, this appears to be isolated to the Azure Databricks Serverless notebook compute path, rather than a Unity Catalog permission or table-level issue.

Since:

• Python execution works normally.
• SELECT 1 and USE CATALOG fail/hang only on Serverless.
• The same Unity Catalog operations work successfully from a SQL Warehouse.
• A fresh notebook and Serverless session produce the same behavior.
• The error is PERMISSION_DENIED: Cert validation failed. Origin workspace.

I would recommend checking the workspace's Serverless networking/security configuration, particularly any serverless network policies, firewall/NSP rules, private connectivity, or other network controls that could affect Serverless compute. Azure Databricks Serverless uses a managed compute plane, and its connectivity to Azure resources can be controlled through Network Connectivity Configurations (NCCs) and serverless network policies.

However, because SELECT 1 itself fails and the SQL Warehouse works, if no customer-side network restriction is identified, this should be escalated to Microsoft/Azure Databricks Support as a possible Serverless backend/control-plane certificate validation issue.

When opening the support request, provide:

1. Workspace URL/ID and Azure region
2. Exact UTC timestamp of the failure
3. The full PERMISSION_DENIED: Cert validation failed. Origin workspace error
4. Confirmation that SQL Warehouse can access the same catalog successfully
5. Confirmation that a new Serverless session and notebook were tested
6. Screenshot/logs showing the failure

Ask support to specifically investigate the Serverless notebook compute → workspace/Unity Catalog service certificate validation path.

This distinction is important because Azure Databricks documentation notes that SSL/certificate failures can be caused by networking/security configuration, while Serverless compute has its own managed networking path

Based on the additional testing, this no longer looks like a Unity Catalog permission issue, catalog binding issue, or problem with the underlying tables.

The key observation is that the same catalog operations (SHOW CATALOGS, USE CATALOG, SHOW SCHEMAS) and table queries execute successfully through a SQL Warehouse, while even a trivial SELECT 1 fails from Serverless notebook compute with:

PERMISSION_DENIED: Cert validation failed. Origin workspace

That points to a failure in the Serverless notebook compute path rather than a problem with the catalog objects themselves.

A few targeted checks that may help further isolate the issue:

Verify whether any serverless network connectivity configuration (NCC) or serverless egress policy is associated with the workspace. Certificate validation failures can occur when Serverless compute cannot successfully reach required control-plane or governance endpoints.

Compare behaviour between Serverless notebook compute and a classic/all-purpose cluster in the same workspace. If Spark SQL succeeds on classic compute but fails only on Serverless, that further narrows the issue to the Serverless execution path.

1. Run a simple Spark operation from Python rather than SQL:

   spark.range(1).count()

If this also hangs or fails, it strengthens the conclusion that the issue occurs when a Spark session attempts to initialise or communicate with backend services, rather than being specific to SQL parsing or Unity Catalog commands.

1. Check whether the issue is reproducible across multiple users in the same workspace. If all users experience the same behaviour, that would indicate a workspace-level Serverless configuration issue rather than a user-specific permission problem.

Given that SQL Warehouse access to the metastore and tables is working correctly, I would avoid spending additional time recreating catalogs, changing table grants, or modifying Unity Catalog permissions. The evidence gathered so far points away from those areas and toward the Serverless notebook compute environment itself.

Help make this community better for everyone: if this answer resolved your issue, please accept it or leave an upvote. If not, share more details in a comment so we can continue the discussion and find the right solution.