Fabric API: Unable to create Azure Databricks connection using Service Principal

Hi ogureisuo,

Thanks for reaching community, will happy to assist.

The error "Kind: AzureDatabricks is not supported" is the key clue here, and it points to a real current limitation of the Fabric Connections API rather than a request body formatting issue.

Root Cause The Actual Problem

The Fabric Core Connections API (/core/connections/create-connection) does not currently support creating Azure Databricks connections programmatically via the gateway path regardless of whether you use "Databricks" or "AzureDatabricks" as the kind. The DMTS_UnsupportedConnectionStringError with Kind: AzureDatabricks is not supported confirms the gateway engine simply doesn't accept this connection type via the API in the way you're attempting.

What Actually Works The Correct Approach

For the Fabric Azure Databricks connector in pipelines and copy activities, Personal Access Token (PAT) is the supported authentication kind not Service Principal directly. Microsoft Learn

The recommended automation path is:

Step 1 Generate a PAT on behalf of your Service Principal

Your Service Principal can generate its own Databricks PAT programmatically:

# Step 1: Get an Entra ID token for your SP
POST https://login.microsoftonline.com/{tenant_id}/oauth2/v2.0/token
Body: client_id, client_secret, scope=2ff814a6-3304-4ab8-85cb-cd0e6f879c1d/.default

# Step 2: Use that token to generate a Databricks PAT
POST https://<databricks-workspace-url>/api/2.0/token/create
Authorization: Bearer <entra_id_token_from_step1>
Body: { "comment": "Fabric automation token", "lifetime_seconds": 7776000 }

Step 2 Create the Fabric connection using PAT

Use the correct request body with PAT auth:

{
  "connectivityType": "ShareableCloud",
  "displayName": "AzureDatabricks_Connection",
  "connectionDetails": {
    "type": "AzureDatabricks",
    "parameters": [
      { "name": "server", "value": "https://<workspace>.azuredatabricks.net" },
      { "name": "httpPath", "value": "/sql/1.0/warehouses/<warehouse_id>" }
    ]
  },
  "credentialDetails": {
    "credentialType": "Key",
    "credentials": {
      "key": "<databricks_PAT_from_step1>"
    },
    "encryptedConnection": "Encrypted",
    "encryptionAlgorithm": "None",
    "privacyLevel": "Organizational",
    "skipTestConnection": false
  }
}

Important Caveats

• For copy activity specifically, Personal Access Token is the only available authentication kind Service Principal is not directly supported as an auth type in the Fabric connector for Databricks. Microsoft Learn
• It is not currently possible to use Microsoft Fabric Workspace Identities to access the Databricks SQL API a service principal granted access to Unity Catalog with a PAT generated on its behalf is the current recommended pattern. Medium
• If you need Service Principal OAuth (not PAT), that path currently only works for Dataflow Gen2 connections, not pipeline copy activities

Summary

Hope this unblocks you! Mark as Accepted Solution if it helps