Skip to content

Redshift

Amazon Redshift.

Requirements

  • A Redshift account
  • Rights to create users and databases in Redshift (i.e. super user account).

Setup

Step 1: Run the setup script in your Redshift Cluster

Run the following commands in your redshfit cluster to create a user and database for Continual to use.

-- create user
create user continual --replace this
password 'REPLACE THIS WITH A SECURE PASSWORD'; --replace this

--create database
CREATE DATABASE
continual; --replace this

--grant access
GRANT CREATE, TEMPORARY ON DATABASE
continual to continual;

Note

The Redshift connector currently does not support spaces or percent signs (%) in passwords.

Step 2: Grant Access to Other Resources to Continual

Any tables that you plan to use as source tables for feature sets or models in Continual need to be accessible by the continual user created above. Please grant SELECT permission on any such resources.

Note

If the user that has been created does not have CREATE SCHEMA privileges then a schema must be created for this user. When creating a Redshift-backed project provide this schema name when configuring the connection.

(Optional) Step 3: Configure Cluster VPC to Allow Continual Access

If your cluster restricts access via IP address, you'll need to configure your Redshift's VPC to allow access from Continual. Please contact us to obtain a stable IP for use.

Step 4: Create a Redshift-backed project

During project creation, select Redshift as your feature store. Use the Redshift Configuration you created in step one to configure the connection.

You're now all set! You will be able to use Continaul to create feature sets and models in Redshift!

Cross database query support

Redshift does currently support cross database queries however there are limitations and requirements.

In particular Redshift only supports this on RA3 node types. Please check the requirements and limitations for Redshift cross database queries before attempting to use.

Redshift limitations

Creating too many views in redshift can lead to issues with unsupported correlated subqueries. It is recommended to query tables as opposed to views when defining feature sets and models.

Back to top