Project Nessie¶

Git-like version control for Iceberg tables with branching and tagging.

Quick Start¶

make example-nessie

Open http://localhost:9091 to access the Floe UI.

Configuration¶

FLOE_CATALOG_TYPE=NESSIE
FLOE_CATALOG_NAME=demo
FLOE_CATALOG_NESSIE_URI=http://nessie:19120/api/v1
FLOE_CATALOG_NESSIE_REF=main
FLOE_CATALOG_WAREHOUSE=s3://warehouse/

# S3/MinIO storage (required for reading table data)
FLOE_CATALOG_S3_ENDPOINT=http://minio:9000
FLOE_CATALOG_S3_ACCESS_KEY_ID=admin
FLOE_CATALOG_S3_SECRET_ACCESS_KEY=password
FLOE_CATALOG_S3_REGION=us-east-1

Options¶

Variable	Required	Description
`FLOE_CATALOG_NESSIE_URI`	Yes	Nessie API endpoint (e.g., `http://nessie:19120/api/v1`)
`FLOE_CATALOG_NESSIE_REF`	No	Branch or tag (default: `main`)
`FLOE_CATALOG_NESSIE_TOKEN`	No	Bearer token for authentication
`FLOE_CATALOG_WAREHOUSE`	Yes	Warehouse location (e.g., `s3://warehouse/`)
`FLOE_CATALOG_S3_ENDPOINT`	Yes*	S3/MinIO endpoint (*required for MinIO or custom S3)
`FLOE_CATALOG_S3_ACCESS_KEY_ID`	Yes	S3 access key
`FLOE_CATALOG_S3_SECRET_ACCESS_KEY`	Yes	S3 secret key
`FLOE_CATALOG_S3_REGION`	No	S3 region (default: `us-east-1`)

Authentication¶

For authenticated Nessie servers, provide a Bearer token:

FLOE_CATALOG_NESSIE_TOKEN=your-jwt-token

Engine Compatibility¶

Both Spark and Trino have full support for all Iceberg maintenance operations with Nessie.

How It Works¶

Nessie stores pointers to Iceberg table metadata (metadata location, snapshot ID, schema ID) rather than the actual data. Table data remains in object storage (S3/MinIO).

Implicit namespaces: Nessie namespaces are derived from table paths and don't need to be explicitly created. When you create a table at test.orders, the test namespace is automatically recognized.