CockroachDB

The CockroachDB plugin discovers databases, tables, views and materialized views from CockroachDB clusters. It captures columns (including computed column expressions and comments), row and column counts, table sizes, partitioning, foreign key relationships and, for each view, a lineage edge from every table it reads to the view.

Tables and views are named database.schema.table, since one cluster holds many databases. Every database except system is discovered by default; set database to discover one, or exclude_databases to skip more. Hidden columns CockroachDB adds on its own (the rowid of a table without a primary key, the shard column of a hash-sharded index) are left out.

Connection

The plugin speaks the PostgreSQL wire protocol over the SQL port (26257 by default). Set ssl_mode to verify-full with ssl_root_cert for secure clusters, and ssl_cert plus ssl_key for certificate authentication. Leave password empty on insecure clusters.

Required Permissions

The user needs to connect to each database and read its tables:

CREATE USER marmot_reader WITH PASSWORD 'your-password';
GRANT CONNECT ON DATABASE your_db TO marmot_reader;
GRANT USAGE ON SCHEMA your_db.public TO marmot_reader;
GRANT SELECT ON ALL TABLES IN SCHEMA your_db.public TO marmot_reader;

Table sizes come from span statistics, which need one extra system privilege. Without it the plugin logs a warning and skips asset.size_bytes:

GRANT SYSTEM VIEWACTIVITYREDACTED TO marmot_reader;

name: my-cockroachdb-pipeline
runs:
  - cockroachdb:
      host: "<host>"
      user: "<user>"

$ marmot ingest -c ingest.yaml