Connecting Databricks on Clarisights

Connect your Databricks lakehouse to Clarisights so first-party tables and views can be read alongside the rest of your marketing and analytics data. Clarisights queries Databricks read-only through a SQL Warehouse, which lets you blend your warehouse data with channel performance, attribution, and finance datasets.

At a glance

Connector type Data warehouse / lakehouse (read-only)

Authentication Personal Access Token (PAT) on a SQL Warehouse

Permissions needed CAN USE on the SQL Warehouse and SELECT on the table or view

Object support Tables and views (Delta tables and Unity Catalog supported; Hive metastore also supported for legacy workspaces)

Refresh cadence Set per pipeline

Limited rollout Yes (contact your Customer Success Manager to enable)

Authentication options

Clarisights authenticates to Databricks using a Personal Access Token (PAT) tied to a SQL Warehouse in your workspace. The PAT is sent on every query and identifies the principal Clarisights runs as.

Strongly recommended: create the PAT against a service principal, not a personal user. Personal PATs break when the user leaves the org or rotates their credentials, which silently stalls the pipeline. A service-principal PAT keeps the connection stable across team changes.
If your workspace enforces IP allow-lists, you will also need to allow Clarisights' outbound IPs (shared during setup).

Setting up the connection

Step 1: Create a service principal in Databricks

In your Databricks workspace, create a service principal that will own the connection to Clarisights. Using a service principal (instead of a personal account) keeps the integration stable when team members change.

For step-by-step instructions on adding a service principal, see Databricks' guide: Manage service principals.

Step 2: Create a Personal Access Token for the service principal

Generate a PAT for the service principal you just created. Copy the token immediately, because Databricks only shows it once. Set an expiry that matches your rotation policy.

⚠ Common error: Authentication failed when Clarisights validates the connection → caused by an invalid, expired, or revoked PAT → fix by generating a fresh PAT for the service principal and resharing it.

Step 3: Grant permissions on the SQL Warehouse and the data

Grant the service principal:

CAN USE on the SQL Warehouse Clarisights will query through.
SELECT on each table or view that should be exposed to Clarisights.

Grant only the objects you want Clarisights to read. The connection cannot reach anything the service principal does not have SELECT on.

Step 4: Share connection details with Clarisights

Send your CSM the values listed in Connection details exchange below. Clarisights will validate the credentials end-to-end before turning the pipeline on.

⚠ Common error: Warehouse not found or HTTP path invalid → caused by a mistyped SQL Warehouse HTTP path → fix by copying the path directly from SQL Warehouses → your warehouse → Connection details in Databricks (it looks like /sql/1.0/warehouses/<id>).

⚠ Common error: First query of the day takes 30 seconds to a couple of minutes → caused by SQL Warehouse cold start → fix by enabling auto-start on the warehouse, or keeping it warm during your pipeline window.

Connection details exchange

You provide to Clarisights Clarisights provides to you

Workspace URL (e.g. dbc-xxxxx.cloud.databricks.com)

Outbound IPs to add to your IP allow-list, if your workspace enforces one

SQL Warehouse HTTP path (e.g. /sql/1.0/warehouses/<id>)

Confirmation that credentials validated and the pipeline is scheduled

Personal Access Token (service principal recommended)

Catalog · schema · table or view name for each object to expose

What we read

Clarisights reads the tables and views you explicitly grant, and nothing else. The connector supports:

Delta tables in your lakehouse.
Views, including views that join or transform underlying tables.
Unity Catalog three-part identifiers (catalog.schema.table).
Hive metastore two-part identifiers (schema.table) for legacy workspaces.

Each Clarisights channel reads one table or view, exactly as it is; we do not sample it or add joins on the Databricks side. To bring in several tables, either expose them through one view or set up one channel per table.

Connector specifics

SQL Warehouse must be available when the pipeline runs. Either keep the warehouse running or enable auto-start. A cold warehouse can add 30 seconds to ~2 minutes to the first query of the day.
Use a service-principal PAT, not a personal PAT. Personal PATs are tied to a user account and stop working when that user leaves, rotates credentials, or loses workspace access. A service principal removes that fragility.
Identifier formats. Unity Catalog uses catalog.schema.table; the legacy Hive metastore uses schema.table. Both are supported, so share the form your workspace uses when sending the table list.

Limitations & known constraints

One object per channel. Each channel reads a single table or view. Put filters, joins, and other transformations into a view on your side.
Query timeouts. Long-running queries are subject to your SQL Warehouse's timeout settings. Heavy transformations should be materialized as a table, or kept in a view that stays fast to query.
Cold-start lag. If the SQL Warehouse is stopped when Clarisights queries it, the first query has to wait for compute to come up. Enable auto-start to avoid pipeline failures during quiet hours.
Read-only. The connector only reads; Clarisights never writes back to your Databricks workspace.

Operating notes

Refresh cadence is set per pipeline during onboarding (e.g. hourly, daily). Tell your CSM the cadence you need; they will tune it against your warehouse capacity.
Rotate the PAT on schedule. When you rotate, share the new token with your CSM before the old one expires to avoid a gap in the pipeline.
Limited rollout. The Databricks connector is in limited rollout. Contact your Customer Success Manager to enable it for your workspace.

Need help?

When contacting support from the in-app messenger, please include:

The integration name and account ID (Integrations → Channel)
The exact error message or screenshot
The step where the issue occurred
When the issue started