How apache/gravitino works
Project type and purpose
• Apache Gravitino™ is a full-stack, geo-distributed metadata lake platform. It comprises a Java-based server exposing REST APIs, a core metadata engine, plugin-based connectors to external metadata sources (file systems, JDBC catalogs, Hive, Iceberg, Paimon, etc.), engine-specific connectors (Spark, Flink, Trino), client libraries (Java, Python), CLI, Web UI, and filesystem integrations (GVFS FUSE, Hadoop).
File structure
• Top-level modules:
• core, common: central dispatchers, entity store, DTOs, utilities
• server-common, server: HTTP (Jetty) server, REST controllers, authentication filters
• web: Next.js-based React UI
• clients: CLI, Java SDK, Python SDK, FUSE and Hadoop FS adapters
• bundles, catalogs, authorizations: SPI-based plugins for credential providers, file-systems (S3, GCS, OSS, ADLS), metadata sources (Hive, MySQL, PostgreSQL, Doris, Kafka, Iceberg, Paimon, Hudi), auth plugins (Ranger, JDBC, chained)
• spark-connector, flink-connector, trino-connector: engine integrations
• lineage, metrics: auxiliary services
• Build scripts (Gradle), configuration templates (conf/), Docker & Helm charts (dev/, dev/docker/, dev/charts/), docs/, integration tests.
README insights
• Purpose: unified metadata management, direct metadata synchronization via connectors, end-to-end governance, multi-engine connectors, geo-distribution.