MongoDB connector

Use mongodb: or mongodb+srv:. MongoDB supports source and sink roles for all three source modes and complete or incremental sink modes. The sink does not support complete-atomic mode. It is a document sink, not a transactional table staging target.

Source selection

source.table names the collection. source.where accepts a BSON filter, source.columns accepts a BSON projection, and source.query accepts a JSON array representing an aggregation pipeline. Parallel reads add skip/limit boundaries, and ordinary collection reads sort by object ID; validate ordering and pipeline cost on the actual collection.

Field mapping and writes

MongoDB-to-MongoDB transfers retain each source document. Relational source rows become documents whose field names are normalized for sink keys and lowercased otherwise. Primitive, temporal, binary, BLOB/CLOB, and PostgreSQL JSON-like values take connector-specific conversion paths, so test nested, null, binary, and mixed-type fields before production.

Sink writes use bulk operations. Incremental merge discovers unique-index fields for identity and rejects staged documents with missing or null merge keys. It does not propagate source deletes. sink.auto.create is a SQL DDL feature and is not supported for MongoDB collections.

SRV discovery, TLS, and authentication are properties of the deployment and connection string. Resolve security through the runtime credential chain and keep it out of committed examples.