Storage Modes

The Analytics Store supports two storage modes for exported data: ducklake (default) and parquet. Both produce Parquet files, but differ in how they manage metadata and catalogs.

DuckLake (Default)

DuckLake-managed Parquet files with a catalog storing metadata. This is the default storage mode.

yaci.store.analytics.storage.type=ducklake

Catalog Options

PostgreSQL catalog (recommended for production):

yaci.store.analytics.ducklake.catalog-type=postgresql

When using PostgreSQL as the catalog, it defaults to the main datasource connection. You can optionally configure a separate catalog connection:

yaci.store.analytics.ducklake.catalog-url=jdbc:postgresql://localhost:5432/yaci_store
yaci.store.analytics.ducklake.catalog-username=postgres
yaci.store.analytics.ducklake.catalog-password=pass

DuckDB file catalog (development / single-instance):

yaci.store.analytics.ducklake.catalog-type=duckdb
yaci.store.analytics.ducklake.catalog-path=./data/analytics/ducklake.catalog.db

Characteristics

• ACID transactions, time-travel queries, schema evolution
• PostgreSQL catalog supports multi-instance deployments
• DuckDB catalog is lightweight for development
• Query tables directly by name (e.g., SELECT * FROM analytics.block)

Parquet

Direct Parquet file export without a catalog. Files are organized by table name and partition (date or epoch).

yaci.store.analytics.storage.type=parquet

Directory Layout

./data/analytics/
  block/
    date=2024-01-15/data.parquet
    date=2024-01-16/data.parquet
  epoch_stake/
    epoch=450/data.parquet

Characteristics

• Simple, no external dependencies beyond DuckDB JDBC
• Files are immutable once written
• No built-in catalog metadata
• Query files using read_parquet() with Hive partitioning

Choosing a Storage Mode