MongoFlink is a connector between MongoDB and Apache Flink. It acts as a Flink sink (and an experimental Flink bounded source), and provides transaction mode(which ensures exactly-once semantics) for MongoDB 4.2 above, and non-transaction mode for MongoDB 3.0 above.

The official Flink MongoDB connector is released, thus MongoFlink would only have bugfix updates and remain as a MongoDB connector for Flink 1.15 or below.

• Flink 1.12 above. MongoFlink is built on top of the new sink API added in Flink 1.12.0.
• MongoDB 3.0 above. The official MongoDB Java driver supports 3.0 above.
• JDK 1.8 above.

For Maven users, add the following dependency in your project's pom.xml.

<dependency>
	<groupId>org.mongoflink</groupId>
	<artifactId>mongo-flink</artifactId>
	<version>1.0-SNAPSHOT</version>
</dependency>

MongoFlink heavily relies on Flink connector interfaces, but Flink interfaces may not have good cross version compatibility, thus it's recommended to choose the version of MongoFlink that matches the version of Flink in your project.

With regard to MongoDB compatibility, please refer to MongoDB's docs about the Java driver.

In case there's no version that fits your need, it's recommended to build your own one. See [Build from source] section below.

Use MongoSink in your Flink DataStream application.

	StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

	// non-transactional sink with a flush strategy of 1000 documents or 10 seconds
	MongoConnectorOptions options = MongoConnectorOptions.builder()
		.withDatabase("my_db")
		.withCollection("my_collection")
		.withConnectString("mongodb://user:[email protected]:27017")
		.withTransactionEnable(false)
		.withFlushOnCheckpoint(false)
		.withFlushSize(1_000L)
		.withFlushInterval(Duration.ofSeconds(10))
		.build();

	env.addSource(...)
	.sinkTo(new MongoSink<>(new StringDocumentSerializer(), options));

	env.execute();

Use MongoSink in your Flink Table/SQL application.

	TableEnvironment env = TableEnvironment.create(EnvironmentSettings.inStreamingMode());

	env.executeSql("create table tbl_user_gold (" +
			"    user_id long," +
			"    gold long," +
			"    PRIMARY key(user_id) NOT ENFORCED" +
			") with (" +
			"    'connector'='mongo'," +
			"    'connect_string' = 'mongodb://user:[email protected]:27017'," +
			"    'database' = 'mydb'," +
			"    'collection' = 'user_gold'" +
			")"

	Table userGold = env.executeQuery("select * from tbl_user_gold");
	);

MongoFlink can be configured using MongoConnectorOptions(recommended) or properties in DataStream API and properties in Table/SQL API.

MongoFlink internally converts row data into bson format internally, so its data type mapping is similar to json format.

Checkout the project, and use Maven to build the project locally.

$ mvn verify

Welcome to file an issue if you need help on adopting MongoFlink. Please describe your environment in the issue (e.g. what MongoDB/Flink version you're using).

Welcome to open pull requests if you want to improve MongoFlink. Please state the purpose (and the design if it's a big feature) for better review experience.