Navigation

Kafka Source Connector Guide

The MongoDB Kafka Source Connector moves data from a MongoDB replica set into a Kafka cluster. The connector configures and consumes change stream event documents and publishes them to a topic.

Change streams, a feature introduced in MongoDB 3.6, generate event documents that contain changes to data stored in MongoDB in real-time and provide guarantees of durability, security, and idempotency. You can configure change streams to observe changes at the collection, database, or deployment level. See An Introduction to Change Streams for more information.

Note

Change streams require a replicaSet or a sharded cluster using replicaSets.

Change Stream Event Document Format

A change stream event document contains several fields that describe the event:

  • The top-level _id field is used as the resume token which is used to start a change stream from a specific point in time.
  • The operationType field identifies the type of change represented in the change stream document. Possible values include: “insert”, “update”, “replace”, “delete”, “invalidate”, “drop”, “dropDatabase”, and “rename”.
  • The fullDocument field contents depend on the operation as follows:
    • For insert and replace operations, it contains the new document being inserted or replacing the existing document.
    • For update operations, it contains the complete document that is being updated at some point in time after the update occurred. If the document was deleted since the update, it contains a null value.
  • The documentKey contains either the _id field of the document that was updated or all the components of a shard key for sharded collections.
  • The txnNumber and lsid identify the transaction if the change occurred within one.
{
  _id: { <BSON Object> },
  "operationType": "<operation>",
  "fullDocument": { <document> },
  "ns": {
    "db": <database>,
    "coll": <collection>
  },
  "to": {
    "db": <database>,
    "coll": <collection>
  },
  "documentKey": {
    _id: <value>
  },
  "updateDescription": {
    "updatedFields": { <document> },
    "removedFields": [ <field>, ... ]
  },
  "clusterTime": <Timestamp>,
  "txnNumber": <NumberLong>,
  "lsid": {
    "id": <UUID>,
    "uid": <BinData>
  }
}

Source Connector Configuration Properties

The MongoDB Kafka Source Connector uses the following settings to create change streams and customize the output to save to the Kafka cluster. For an example source connector configuration file, see MongoSourceConnector.properties.

Name Type Description
connection.uri string

Avoid Exposing Your Authentication Credentials

To avoid exposing your authentication credentials in your connection.uri setting, use a ConfigProvider and set the appropriate configuration parameters.

Default: mongodb://localhost:27017,localhost:27018,localhost:27019
Accepted Values: A valid MongoDB connection URI string
database string
Name of the database to watch for changes. If not set, all databases are watched.
Default: “”
Accepted Values: A single database name
collection string
Name of the collection in the database to watch for changes.
The collection in the database to watch. If not set then all collections will be watched.
Default: “”
Accepted Values: A single collection name
publish.full.document.only boolean
Only publish the changed document instead of the full change stream document. Sets the change.stream.full.document=updateLookup automatically so updated documents will be included.
Default: false
Accepted Values: true or false
pipeline string
An array of objects describing the pipeline operations to run.

Example

[{"$match": {"operationType": "insert"}}, {"$addFields": {"Kafka": "Rules!"}}]
Default: []
Accepted Values: A valid JSON array
collation string
A JSON collation document that contains options to use for the change stream. Append .asDocument().toJson() to the collation document to create the JSON representation.
Default: “”
Accepted Values: A valid JSON document representing a collection
batch.size int
The cursor batch size.
Default: 0
Accepted Values: An integer
change.stream.full.document string
Determines what to return for update operations when using a Change Stream. When set to ‘updateLookup’, the change stream for partial updates will include both a delta describing the changes to the document as well as a copy of the entire document that was changed from some point in time after the change occurred.
Default: “”
Accepted Values: “” or default or updateLookup
poll.await.time.ms long
The amount of time to wait before checking for new results on the change stream
Default: 5000
Accepted Values: An integer
poll.max.batch.size int
Maximum number of change stream documents to include in a single batch when polling for new data. This setting can be used to limit the amount of data buffered internally in the connector.
Default: 1000
Accepted Values: An integer
topic.prefix string
Prefix to prepend to database & collection names to generate the name of the Kafka topic to publish data to.
Default: “”
Accepted Values: A string
copy.existing boolean
Copy existing data from source collections and convert them to Change Stream events on their respective topics. Any changes to the data that occur during the copy process are applied once the copy is completed.
Default: false
Accepted Values: true or false
copy.existing.max.threads int
The number of threads to use when performing the data copy. Defaults to the number of processors.
Default: defaults to the number of processors
Accepted Values: An integer
copy.existing.queue.size int
The max size of the queue to use when copying data.
Default: 16000
Accepted Values: An integer

Note

The default maximum size for Kafka messages is 1MB. Update the following Kafka (versions 0.11.0 through 2.2) configuration properties to enable a larger maximum size if the JSON string size of the change stream documents exceeds the maximum:

System Property Name Description
Consumer max.partition.fetch.bytes Maximum size of a message that can be fetched by a consumer.
Broker replica.fetch.max.bytes Maximum size of a message that can be replicated within a Kafka cluster.
Broker message.max.bytes Maximum size of a message from a producer that is accepted by the broker.
Producer max.message.bytes Per referenced topic, the maximum size of an uncompressed message that can be appended to a topic.

Custom Pipeline Example

You can use the pipeline configuration setting to define a custom aggregation pipeline to filter or modify the change events output. In this example, we set the pipeline configuration to observe only insert change events:

pipeline=[{"$match": {"operationType": "insert"}}]

Note

Make sure the results of the aggregation pipeline contain the top-level _id field which MongoDB uses as the resume token.

Topic Naming Example

The MongoDB Kafka Source connector publishes the changed data events to a Kafka topic that consists of the database and collection name from which the change originated. For example, if an insert was performed on the test database and data collection, the connector will publish the data to a topic named test.data.

If the topic.prefix configuration is set to true, the Kafka topic name will be prepended with the specified value. For example:

topic.prefix=mongo

Once set, any data changes to the data collection in the test database are published to a topic named mongo.test.data.

Existing Data Copy Example

The MongoDB Kafka Source connector can be configured to copy existing data from collections in a database to their associated topic as insert events prior to broadcasting change stream events. The connector does not support renaming a collection during the copy process.

Data Copy Can Produce Duplicate Events

If clients make changes to the data in the database while the source connector is converting existing data, the subsequent change stream events may contain duplicates. Since change stream events are idempotent, the data is eventually consistent.

The following configuration example instructs the connector to copy all collections in the example database, convert the data to change stream events, and broadcast changes after any collection data is updated.

database=example
copy.existing=true

How to Recover from An Invalid Resume Token

In the event your Connector pauses or shuts down long enough for the Change Stream resume token to expire from the MongoDB Oplog, you may encounter an error that prevents you from starting up the Connector. If you encounter this condition, you must delete the topic data referenced by the offset.storage.topic setting if you are using Distributed Mode or the file referenced by the offset.storage.file.filename setting (e.g. /tmp/connect.offsets) if you are using the Standalone mode. After you delete the appropriate data, you should be able to start your Connector workers and listen to the change stream.