Apache Iceberg
Connect an Apache Iceberg table to Lakehousecat either directly by its metadata location in S3-compatible object storage, or through a REST catalog (e.g. Nessie, Polaris, Lakekeeper).
Connection Fields
| Field | Required | Description |
|---|---|---|
| URL | Yes (metadata path) | Full URL of the Iceberg table in object storage. Example: http://seaweedfs-s3:8333/iceberg/tablename |
| Access Key ID | Yes | The access key for object storage authentication |
| Secret Access Key | Yes | The secret key for object storage authentication |
| S3 Endpoint | No | Custom S3-compatible endpoint address (MinIO, Ceph, Wasabi, Hetzner, etc.). Leave empty for AWS. Applies to both the metadata path and the catalog path. |
| Catalog URI | No | REST catalog endpoint, e.g. http://nessie-host:19120/iceberg/main. Include the branch/ref in the path where the catalog supports it (Nessie: /iceberg/<branch>). Leave empty to use the metadata path instead. |
| Warehouse | No | Warehouse identifier as configured on the catalog server. Only used when Catalog URI is set. |
| Token | No | Bearer token for catalog authentication, where required. Masked after saving. |
Two Ways to Connect
- Metadata path (default): point directly at the table's metadata directory in object storage. Works with any Hadoop-style warehouse, no catalog infrastructure needed.
- Catalog path: set Catalog URI (and optionally Warehouse/Token) to resolve
tables through a REST catalog instead. Namespaces are folded into the table name (e.g.
lhc.customers→lhc__customers); the catalog is the source of truth for table identity and the current snapshot, not the file layout.
Prerequisites
- Admin or Builder role in Lakehousecat
- An Apache Iceberg table accessible via an S3-compatible endpoint, or a reachable REST catalog
- Object storage credentials with read access to the table path
- For the catalog path: a catalog token if the catalog requires authentication
Notes
- The URL (metadata path) must point directly to the Iceberg table root directory (the folder containing the metadata).
- Network connectivity from the Lakehousecat backend to the object storage endpoint — and, when used, the catalog endpoint — is required.
- The metadata path and catalog path are independent; setting a Catalog URI does not remove the need for object storage credentials, since table data still lives in S3-compatible storage.
- AWS Glue is not supported as a catalog type yet.
Next Steps
After creating the datasource, go to the Operations tab to trigger semantic extraction.