Skip to content

FeatureByte to Data Warehouse Authentication

Initial Setup to Data Warehouse

FeatureByte operates by pushing all feature computation into a schema in your data warehouse. Hence, it is crucial that FeatureByte is provided with a schema and a credentials that has sufficient permissions to create tables, views and run queries in that schema.

For data sources locations, please provide read-only access to the credential. FeatureByte is designed to never modify source tables to ensure data integrity.

Note: For the schema that is given to FeatureByte. Do not place any source tables in that schema. FeatureByte is designed to mitigate circular data sourcing by ensuring that source tables are never in the same schema as the tables and views created by FeatureByte. Any tables or views present in that schema is ignored and can not be used as source tables.

Credentials Management

In featurebyte we access your data warehouse using a service account token. Depending on your data warehouse, the authentication method may differ.

In simple terms there are 2 ways that a user of FeatureByte platform can authenticate to the data warehouse. The user can manually input their personal api token or they can be part of a featurebyte group which has a service account token binded to it.

Having a featurebyte group credential is recommended as it simplifies the management of credentials, e.g. rotating the api token is as simple as changing 1 token instead of changing tokens for every user. However, as the token is shared among the group, you will be unable to track individual user actions on the data warehouse. If you need to track individual user actions, you will need to use personal credentials.

Special Note: When you first bind a data warehouse to featurebyte, that credential must be kept up to date and valid. It will be used to apply any warehouse migration scripts during featurebyte upgrades.

Personal Credentials

Personal data warehouse credentials can be added/updated in the > Security > Feature Store Credential page.

Workflow Diagram

Shared Credentials

Create a featurebyte group in the > Admin > User Group page.

Workflow Diagram

Bind a credential to the user group at the following page > Security > Feature Store Credential page.

Workflow Diagram

Now when a user logs in, any request to the data warehouse will be authenticated as the credential binded to the group.

Validation

You can validate that the authentication is working by going to the > Explore page. Select the featurestore you would want to view the tables of and see that the tables are listed.

Troubleshooting

There are a few common reasons why you may not be able to see the tables in the explore page.

  1. The user does not have a personal data warehouse credential binded to the data warehouse.
  2. The user is not part of a featurebyte group that has a credential binded to that particular data warehouse.
  3. The credential (personal or group) is invalid or has expired.
  4. Networking errors, e.g. featurebyte server cannot reach the data warehouse.
  5. The credential does not have sufficient permissions to list the tables in the data warehouse.
  6. The order of credential binding goes by personal credential first then newest group credential binded to the user. Hence, there is a case where a user has a personal credential binded to the data warehouse but it is invalid, and the user is part of a group that has a valid credential binded to the data warehouse. In that case, please remove the personal credential binded to the user.