Skip to main content

GCP GCS

You'll create a storage credential and external location to connect Databricks to GCS in ~10 min.

Prereqs: Cloud Object Storage overview, GCP project with Owner permissions

What you'll build​

A Google Cloud Storage bucket Databricks can read through a Unity Catalog storage credential and external location.

Prerequisites​

  • A GCP project with Owner or Storage Admin permissions.
  • A Databricks workspace with metastore-admin or account-admin privileges.

Steps​

1. Follow the official guide​

GCS is the one cloud where the Databricks docs already do the whole job end to end, so this page hands you off rather than duplicating it:

Connect to a Google Cloud Storage (GCS) external location

The guide creates a service account, grants it access to your bucket, registers the storage credential, and creates the external location.

2. Mark the external location as read-only​

Once the external location exists, set it to read-only. Now no Databricks workload can write to that path, so a stray job cannot clobber your source data.

Follow the guide: Mark an external location as read-only.

warning

Skip this and the location stays writable for any principal with write grants. Leave it read-only unless a pipeline genuinely needs to write back to this path.

Verify​

  1. In the Databricks workspace, navigate to Catalog > External Data > External Locations.
  2. Click the new external location and click Test Connection.
  3. Confirm the test returns a success status.

Where people trip​

Test Connection fails with permission denied

The service account behind the storage credential needs the storage.objectAdmin role (or equivalent) on the target bucket. Grant it and test again.

Storage credential creation fails

Check the service account key or workload identity federation setup. The service account has to live in the same GCP project as the Databricks workspace.

Next​