Apache Gravitino

Apache Gravitino

Apache Gravitino (formerly Datastrato Gravitino) is an open-source metadata management platform that provides a unified catalog abstraction over heterogeneous data sources — Hive Metastore, Iceberg REST Catalog, JDBC databases, cloud storage file systems, and more. Gravitino presents a single API and management plane for metadata across all of these, making it a “meta-catalog” or metadata hub.

Gravitino was accepted into the Apache Software Foundation incubator and is rapidly gaining adoption as a vendor-neutral metadata layer for multi-source data architectures.

What Gravitino Provides

Unified Catalog API

Gravitino exposes a single REST API for creating, managing, and querying metadata across multiple catalog backends:

Gravitino Server
  ├── Iceberg REST Catalog (manages Iceberg tables)
  ├── Hive catalog (connects to existing HMS)
  ├── JDBC catalog (connects to PostgreSQL, MySQL)
  └── Fileset catalog (manages unstructured data in cloud storage)

A single Gravitino API call can create a namespace that spans multiple catalog backends, or provide a unified listing of all tables regardless of their storage format.

Iceberg REST Catalog Compatibility

Gravitino implements the Iceberg REST Catalog specification, allowing Iceberg-compatible engines (Spark, Flink, Trino, Dremio, PyIceberg) to connect to Gravitino and query its managed Iceberg catalogs:

# PyIceberg: connect to Gravitino's Iceberg REST endpoint
from pyiceberg.catalog import load_catalog

catalog = load_catalog(
    "gravitino",
    **{
        "type": "rest",
        "uri": "http://gravitino-host:9001/api/iceberg/metalakes/my-metalake/catalogs/iceberg-catalog",
        "warehouse": "my-metalake/iceberg-catalog",
    }
)

# Standard Iceberg operations work transparently
tables = catalog.list_tables("analytics")
table = catalog.load_table("analytics.orders")

Gravitino Metalake

The top-level organizational unit in Gravitino is the metalake — a tenant boundary that contains multiple catalogs. A Gravitino deployment can serve multiple metalakes for different organizations or environments:

Gravitino Server
  ├── Metalake: production
  │   ├── Iceberg catalog: lakehouse
  │   ├── Hive catalog: legacy_hive
  │   └── JDBC catalog: operational_db
  └── Metalake: development
      ├── Iceberg catalog: dev_lakehouse
      └── JDBC catalog: dev_db

Connecting Gravitino to Iceberg Storage

# Gravitino configuration: creating an Iceberg catalog
catalogs:
  - name: lakehouse
    type: lakehouse-iceberg
    provider: lakehouse-iceberg
    properties:
      catalog-backend: rest # or hive, jdbc
      uri: https://my-polaris.example.com
      warehouse: s3://my-bucket/warehouse/

Or using the Gravitino REST API:

curl -X POST http://gravitino-host:9001/api/metalakes/production/catalogs \
  -H "Content-Type: application/json" \
  -d '{
    "name": "lakehouse",
    "type": "RELATIONAL",
    "provider": "lakehouse-iceberg",
    "properties": {
      "catalog-backend": "hive",
      "uri": "thrift://hms-host:9083",
      "warehouse": "s3://my-bucket/warehouse/"
    }
  }'

Gravitino Use Cases

Legacy Hive + Modern Iceberg Unification

Organizations with both legacy HMS-managed Hive tables and new Iceberg tables can present both through a single Gravitino API:

# Both Hive and Iceberg tables visible through Gravitino
gravitino_catalog = load_catalog("gravitino", ...)

# List all tables (Hive and Iceberg)
all_tables = gravitino_catalog.list_tables("analytics")
for table in all_tables:
    print(table)  # includes both Hive and Iceberg tables

Multi-Format Table Discovery

Data engineers can use Gravitino as a discovery tool to find tables regardless of format, then route queries to the appropriate engine.

Governance Layer

Gravitino’s centralized metadata view enables governance teams to see all tables, schemas, and properties across all catalog types — without needing to connect to each catalog separately.

Gravitino vs. Apache Polaris

Gravitino is well-suited for organizations that need a unified metadata view over diverse, heterogeneous data sources including Iceberg. Polaris is the right choice for organizations building a pure Iceberg lakehouse that needs full credential vending, fine-grained RBAC, and maximum performance.