Lakekeeper
Apache Iceberg REST Catalog implementation for managing data lake tables:
- Iceberg REST Catalog: Complete Apache Iceberg REST specification implementation
- OIDC Authentication: Integrated with Keycloak for secure access via PKCE flow
- PostgreSQL Backend: Reliable metadata storage with automatic migrations
- Web UI: Built-in web interface for catalog management
- Secrets Management: Vault/External Secrets integration for secure credentials
- Multi-table Format: Primarily designed for Apache Iceberg with extensibility
Installation
just lakekeeper::install
During installation, you will be prompted for:
- Lakekeeper host (FQDN): The domain name for accessing Lakekeeper (e.g.,
lakekeeper.yourdomain.com)
The installation automatically:
- Creates PostgreSQL database and user
- Stores credentials in Vault (if External Secrets is available) or Kubernetes Secrets
- Creates Keycloak OIDC client with PKCE flow for Web UI authentication
- Creates API client (
lakekeeper-api) for programmatic access with OAuth2 Client Credentials Flow - Configures audience mapper for JWT tokens
- Runs database migrations
- Configures Traefik ingress with TLS
IMPORTANT: During installation, API client credentials will be displayed. Save these for programmatic access (dlt, PyIceberg, etc.).
Access
Access Lakekeeper at https://lakekeeper.yourdomain.com and authenticate via Keycloak.
Warehouse Management
Creating Warehouses with Vended Credentials
Create warehouses with STS (Security Token Service) enabled for automatic temporary credential management:
# Create warehouse with default name and bucket
just lakekeeper::create-warehouse <warehouse-name> <bucket-name>
# Example: Create 'production' warehouse using 'warehouse' bucket
just lakekeeper::create-warehouse production warehouse
This creates a warehouse with:
- STS enabled for vended credentials (temporary S3 tokens)
- S3-compatible storage (MinIO) with path-style access
- Automatic credential rotation via MinIO STS
Prerequisites:
- MinIO bucket must exist (create with
just minio::create-bucket <bucket-name>) - API client credentials must be available in Vault
Benefits of Vended Credentials:
- No need to distribute static S3 credentials to clients
- Automatic credential expiration and rotation
- Better security through temporary tokens
- Centralized credential management
Creating Namespaces
Namespaces organize tables within a warehouse (similar to databases in traditional systems):
# Create Iceberg namespace in a warehouse
just lakekeeper::create-warehouse-namespace <warehouse-name> <namespace>
# Example: Create 'ecommerce' namespace in 'test' warehouse
just lakekeeper::create-warehouse-namespace test ecommerce
Managing Warehouses
List, view, and delete warehouses:
# List all warehouses
just lakekeeper::list-warehouses
# List all namespaces in a warehouse
just lakekeeper::list-warehouse-namespaces <warehouse-name>
# Example: List namespaces in 'test' warehouse
just lakekeeper::list-warehouse-namespaces test
# Delete a namespace from a warehouse (recursively deletes all tables)
just lakekeeper::delete-warehouse-namespace <warehouse-name> <namespace>
# Example: Delete 'ecommerce' namespace from 'test' warehouse (including all tables)
just lakekeeper::delete-warehouse-namespace test ecommerce
# Delete a warehouse (must be empty)
just lakekeeper::delete-warehouse <warehouse-name>
# Force delete a warehouse (automatically deletes all namespaces first)
just lakekeeper::delete-warehouse <warehouse-name> true
# Example: Force delete 'test' warehouse with all its namespaces
just lakekeeper::delete-warehouse test true
Important Notes:
- Namespace deletion is recursive - it will delete all tables and data within the namespace
- Warehouses must be empty before deletion. If a warehouse contains namespaces, you must either:
- Delete each namespace individually using
delete-warehouse-namespace, then delete the warehouse - Use force deletion (
delete-warehouse <name> true) to automatically delete all namespaces and their tables first
- Delete each namespace individually using
- All deletion operations require confirmation prompts to prevent accidental data loss
Programmatic Access
API Client Credentials
During installation, a default API client lakekeeper-api is automatically created for programmatic access (dlt, Python scripts, etc.).
IMPORTANT: The client ID and secret are displayed during installation. Save these credentials securely.
If you need additional API clients or lost the credentials:
# Create additional API client with custom name
just lakekeeper::create-oidc-api-client my-app
# Recreate default client (delete first, then create)
just lakekeeper::delete-oidc-api-client lakekeeper-api
just lakekeeper::create-oidc-api-client lakekeeper-api
Each API client has:
- Service account enabled for OAuth2 Client Credentials Flow
lakekeeperscope with audience mapper (aud: lakekeeper)- Client credentials stored in Vault (if External Secrets is available)
Using API Clients
dlt (Data Load Tool)
Configure dlt to use the API client credentials:
export OIDC_CLIENT_ID=lakekeeper-api
export OIDC_CLIENT_SECRET=<secret-from-creation>
export ICEBERG_CATALOG_URL=http://lakekeeper.lakekeeper.svc.cluster.local:8181/catalog
export ICEBERG_WAREHOUSE=test # Use warehouse with vended credentials enabled
export KEYCLOAK_TOKEN_URL=https://auth.example.com/realms/buunstack/protocol/openid-connect/token
export OAUTH2_SCOPE=lakekeeper # Optional, defaults to "lakekeeper"
The dlt Iceberg REST destination automatically uses these credentials for OAuth2 authentication and receives temporary S3 credentials via STS (vended credentials).
Notes:
KEYCLOAK_TOKEN_URLis required because Lakekeeper v0.9.x uses external OAuth2 provider (Keycloak) instead of the deprecated/v1/oauth/tokensendpoint.OAUTH2_SCOPEmust be set tolakekeeper(default) to include the audience claim in JWT tokens. PyIceberg defaults tocatalogscope, which is not valid for Keycloak.- No S3 credentials needed when using warehouses with vended credentials enabled (STS). Lakekeeper provides temporary S3 credentials automatically.
Legacy Mode: Static S3 Credentials
If using a warehouse with vended-credentials-enabled=false, you need to provide static S3 credentials:
# Additional environment variables for static credentials mode
export S3_ENDPOINT_URL=http://minio.minio.svc.cluster.local:9000
export S3_ACCESS_KEY_ID=<minio-access-key>
export S3_SECRET_ACCESS_KEY=<minio-secret-key>
To get MinIO credentials:
just vault::get minio/dlt access_key
just vault::get minio/dlt secret_key
Or create a dedicated MinIO user:
just minio::create-user dlt "dlt-data"
PyIceberg
With vended credentials (recommended):
from pyiceberg.catalog import load_catalog
catalog = load_catalog(
"rest_catalog",
**{
"uri": "http://lakekeeper.lakekeeper.svc.cluster.local:8181/catalog",
"warehouse": "test", # Use warehouse with vended credentials enabled
"credential": f"{client_id}:{client_secret}", # OAuth2 format
"oauth2-server-uri": "https://auth.example.com/realms/buunstack/protocol/openid-connect/token",
"scope": "lakekeeper", # Required for Keycloak (PyIceberg defaults to "catalog")
}
)
With static S3 credentials (legacy mode):
catalog = load_catalog(
"rest_catalog",
**{
"uri": "http://lakekeeper.lakekeeper.svc.cluster.local:8181/catalog",
"warehouse": "default",
"credential": f"{client_id}:{client_secret}",
"oauth2-server-uri": "https://auth.example.com/realms/buunstack/protocol/openid-connect/token",
"scope": "lakekeeper",
# Static S3 credentials (only needed when vended credentials disabled)
"s3.endpoint": "http://minio.minio.svc.cluster.local:9000",
"s3.access-key-id": "<minio-access-key>",
"s3.secret-access-key": "<minio-secret-key>",
"s3.path-style-access": "true",
}
)
Trino Integration
Trino uses its own OIDC client with service account. This is automatically configured by just trino::enable-iceberg-catalog. You don't need to create a separate API client for Trino.
Deleting API Clients
# Delete default API client
just lakekeeper::delete-oidc-api-client
# Delete custom-named client
just lakekeeper::delete-oidc-api-client my-app
This removes the Keycloak client and Vault credentials.
Cleanup
To remove all Lakekeeper resources and secrets from Vault:
just lakekeeper::cleanup
This will prompt for confirmation before deleting:
- PostgreSQL database
- Vault secrets
- Keycloak client
Uninstallation
# Keep database
just lakekeeper::uninstall false
# Delete database as well
just lakekeeper::uninstall true
This will:
- Uninstall the Lakekeeper Helm release
- Delete Kubernetes secrets
- Optionally delete PostgreSQL database
- Remove Keycloak OIDC client
Documentation
For more information, see the official documentation: