Query Iceberg tables in AWS Glue, BigLake, OneLake, Polaris, Unity Catalog or any Iceberg REST catalog in place, without copying them.
Connect a catalog and Contrast mounts it as a read-only database. Its namespaces are the database's schemas, so
SELECT * FROM lake.sales.orders reads sales.orders in the catalog as it is when the query runs. Lakeserve reads
the table's files from your storage; nothing is copied into Contrast, and Contrast never writes to your catalog or
storage.
Workspace owners and admins open Sources → Direct → Data catalogs → Add a catalog, choose the catalog, fill in its connection parameters and name the database. Contrast checks the connection the way queries will use it before saving anything: it signs in, lists namespaces and tables, loads a table and reads its metadata file with the storage credentials a query would get. A failed check saves nothing and says what to fix.
| Catalog | Parameters | How table files are read |
|---|---|---|
| AWS Glue | Region; an access key, or an IAM role | The same AWS identity |
| BigLake Metastore | Google ADC client ID, client secret and refresh token, quota project, Cloud Storage bucket | Credentials BigLake vends, else the same Google identity |
| Microsoft OneLake | Fabric workspace ID, lakehouse (data item) ID, Entra tenant ID, application ID and secret | The same Entra application |
| Polaris (Snowflake Open Catalog) | Account host, catalog name, principal client ID and secret, optional storage endpoint | Credentials the catalog vends |
| REST Catalog | Catalog URL, warehouse, optional storage endpoint; OAuth client credentials or a bearer token | Credentials the catalog vends |
| Unity Catalog | Workspace URL, catalog name, service principal client ID and OAuth secret | Credentials Unity Catalog vends |
Choose AWS IAM role and add the principal and external ID the form shows to your role's trust policy. The
external ID is unique to your workspace. The role needs glue:GetCatalog, glue:GetDatabase(s) and
glue:GetTable(s) on the databases and tables to expose, and s3:GetObject and s3:ListBucket on their storage.
Create a service principal and an OAuth secret for it (a secret scoped to unity-catalog read access is enough).
Grant it USE CATALOG on the catalog and USE SCHEMA, SELECT and EXTERNAL USE SCHEMA on each schema, and turn
on External data access for the metastore. Tables in Databricks default storage (serverless workspaces) cannot be
read by any external engine; keep the catalog's managed location on your own cloud storage. Contrast reads the
catalog's Iceberg tables.
The principal's role needs read privileges on the catalog, and the catalog must vend storage credentials
(X-Iceberg-Access-Delegation: vended-credentials). A storage endpoint, when given, limits reads to tables under it.
SHOW SCHEMAS IN lake;
SHOW TABLES IN lake.sales;
SELECT customer, sum(amount) FROM lake.sales.orders GROUP BY customer;Names are case-insensitive and are matched to the catalog's spelling. Quote names with other characters, such as
lake.`sales-eu`.orders, and a nested namespace as one name, lake.`sales.eu`.orders. Each query reads every
table's current snapshot, so new commits are visible immediately. Connected catalogs are read-only: writes, ingestion
into them and table maintenance stay with the catalog's owner.
Disconnect removes the database and its stored credentials. Your catalog and data are untouched.