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.
version | flink version | mongodb driver version |
---|---|---|
0.1 | 1.13.x | 4.2.x |
0.2 | 1.14.x | 4.4.x |
0.3 | 1.15.x | 4.4.x |
1.0.x | 1.15.x | 4.7.x |
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.
option | properties key | description | default value |
---|---|---|---|
MongoConnectorOptions.transactionEnable | sink.transaction.enable | Whether to use transactions in MongoSink (requires MongoDB 4.2+). | false |
MongoConnectorOptions.flushOnCheckpoint | sink.flush.on-checkpoint | Whether to flush the buffered documents on checkpoint barriers. | false |
MongoConnectorOptions.flushSize | sink.flush.size | Max buffered documents before flush. Only valid when sink.flush.on-checkpoint is false . |
1000 |
MongoConnectorOptions.flushInterva | sink.flush.interval | Flush interval in milliseconds. Only valid when sink.flush.on-checkpoint is false . |
30000 |
MongoConnectorOptions.maxInFlightFlushes | sink.max.in-flight.flushes | Max in-flight flushes before blocking the writer. | 5 |
MongoConnectorOptions.upsertEnable | (true if the table has primary keys) | Whether to write documents via upsert mode. | false |
MongoConnectorOptions.upsertKey | (derived automatically from the primary key) | The primary keys for upsert. Only valid in upsert mode. Keys are in csv format for properties. | [] |
MongoFlink internally converts row data into bson format internally, so its data type mapping is similar to json format.
Flink SQL Data Type | MongoDB Data Type |
---|---|
NULL | BsonNull |
BOOLEAN | BsonBoolean |
STRING/VARCHAR/CHAR | BsonString |
INTEGER/SMALLINT/TINYINT | BsonInt32 |
BIGINT/INTERVAL_DAY_TIME | BsonInt64 |
FLOAT/DOUBLE | BsonDouble |
BINARY/VARBINARY | BsonBinary |
DATE | BsonInt32 |
TIMESTAMP(3) | BsonDateTime |
TIMESTAMP_LTZ(3) | BsonDateTime |
DECIMAL | BsonDecimal128 |
ARRAY | BsonArray |
MAP | BsonDocument |
ROW | BsonDocument |
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.