Lakegres

Lakegres Clusters expose a PostgreSQL JDBC endpoint that you can connect to from any PostgreSQL-compatible client. Queries run on Spark under the hood — see Spark SQL syntax for the query dialect.

Catalog and Access Control Modes

A Lakegres Cluster runs in one of three modes. The mode determines how you authenticate and which driver you need.

Mode	Authentication	Authorization	Driver
Onehouse Catalog	Username/password from the cluster connection details page	Onehouse EKS node role	Standard PostgreSQL JDBC driver
Glue Catalog	Username/password from the cluster connection details page	Onehouse EKS node role	Standard PostgreSQL JDBC driver
Glue Catalog with Lake Formation	Corporate identity provider (IdP) via browser OAuth	Lake Formation	Onehouse Lakegres JDBC driver

See Get Connection Details for where to find the endpoint and credentials, and Lakegres JDBC Driver: Identity Provider Integration for the full list of supported IdPs and their driver properties.

Connecting with Lake Formation

When using Lake Formation for access control, you can optionally specify an IAM role to assume for querying. If you don't specify a role, the first IAM role from your authentication token is used by default.

To specify a different IAM role, add it to the connection using the options parameter:

jdbc:postgresql://host:5432/postgres?options=-c onehouse.lakebase_iam_role=arn:aws:iam::account:role/role-name

Client Integrations

• Connect with DBeaver
• Connect from a Java client
• Connect from a Python client
• Connect from psql

warning

If you are using a tool not listed here, please contact Onehouse support for assistance.

Network Access

By default, Lakegres Clusters only accept traffic from within your VPC. To connect from outside the VPC (e.g., your laptop), follow Connect to your VPC to set up a VPN or bastion host.

Troubleshooting connection issues

Endpoint URL changed without warning

Lakegres Clusters can be rebalanced between underlying instances when the engine pool scales — for example, after a min/max engine adjustment. Rebalancing rotates the endpoint host, so clients pinned to a stale hostname will fail to connect.

When connections start failing:

1. Open the cluster on the Clusters page and copy the current endpoint from the connection details panel.
2. Update your client connection strings with the new endpoint.
3. Restart any long-running engines or query tools that cached the previous hostname.

Credentials missing or "queries not running"

Credentials are regenerated on the cluster connection details page when needed. If your stored credentials suddenly stop working, fetch the latest username/password from the cluster page rather than assuming they have been revoked.

If the cluster itself has lost engine capacity (all engines unhealthy), no credentials will work until the engines recover. Check the cluster status in the console; contact support if engines stay unhealthy for more than a few minutes.