DataStream Connectors#

Use these pages for Java builder APIs, connector-specific serialization and deserialization schemas, and the runtime behavior that the Table connectors reuse. For DDL, SQL type mappings, metadata columns, and planner-specific restrictions, see the Table API connectors. The connector overview explains how to choose between the two APIs.

Capability map#

The source column names how Flink distributes reads, while the destination column names the record-level routing surface of the sink. The delivery column summarizes connector-to-service writes; the delivery guarantees page defines the checkpoint and state qualifications. Not applicable means that the capability does not fit the service’s role; it is not an implementation-status claim.

Connector pagesPrimary roleSource and split strategySink deliveryPer-record destinationCDC
BigQuery: Quickstart, DataStream examples source and sink, DataStream reference source and sink, Table reference source, sink, and CDCAnalytics warehouseBounded; Storage Read API stream splitsAt-least-once or exactly-once service writesTable through a resolverUpsert/delete sink on the at-least-once default stream; Experimental (#706)
Cloud Pub/Sub: Quickstart, Examples, DataStream (source, sink), Table (source, sink)MessagingUnbounded; subscription splitsAt-least-onceTopic through a resolverNo native DataStream changelog; Table formats can carry one
Cloud Tasks: Quickstart, Examples, DataStream, TableTask deliveryNot applicable; sink onlyAt-least-onceQueue through a resolverNot applicable
Bigtable: Quickstart, Examples, DataStream source, sink, and Change Streams, Table source, sink, lookup, and Change StreamsWide-column storeBounded; sampled row-range splitsAt-least-onceTable through a resolverChange Streams partitions; unbounded or bounded
Spanner: Quickstart, Examples, DataStream, TableRelational databaseBounded; service-planned partitionsAt-least-onceTable in each serialized mutationChange Streams partitions; unbounded