Pipelining with Kafka Connect and Kafka Streams

Overview

This example shows users how to build pipelines with Apache Kafka®.

It showcases different ways to produce data to Apache Kafka® topics, with and without Kafka Connect, and various ways to serialize it for the Kafka Streams API and ksqlDB.

Example	Produce to Kafka Topic	Key	Value	Stream Processing
Confluent CLI Producer with String	CLI	String	String	Kafka Streams
JDBC source connector with JSON	JDBC with SMT to add key	Long	Json	Kafka Streams
JDBC source connector with SpecificAvro	JDBC with SMT to set namespace	null	SpecificAvro	Kafka Streams
JDBC source connector with GenericAvro	JDBC	null	GenericAvro	Kafka Streams
Java producer with SpecificAvro	Producer	Long	SpecificAvro	Kafka Streams
JDBC source connector with Avro	JDBC	Long	Avro	ksqlDB

Detailed walk-thru of this example is available in the whitepaper Kafka Serialization and Deserialization (SerDes) Examples and the blog post Building a Real-Time Streaming ETL Pipeline in 20 Minutes

Confluent Cloud

You can apply the same concepts explained in this example to Confluent Cloud. Confluent Cloud also has fully managed connectors that you can use, instead of self-managing your own, so that you can run 100% in the cloud. To try it out, create your own Confluent Cloud instance (see the ccloud-stack utility for Confluent Cloud for an easy way to spin up a new environment), deploy a connector, and then point your applications to Confluent Cloud.

Description of Data

The original data is a table of locations that resembles this.

id|name|sale

1|Raleigh|300
2|Dusseldorf|100
1|Raleigh|600
3|Moscow|800
4|Sydney|200
2|Dusseldorf|400
5|Chennai|400
3|Moscow|100
3|Moscow|200
1|Raleigh|700

It produces records to a Kafka topic:

The actual client application uses the methods count and sum to process this data, grouped by each city.

The output of count is:

1|Raleigh|3
2|Dusseldorf|2
3|Moscow|3
4|Sydney|1
5|Chennai|1

The output of sum is:

1|Raleigh|1600
2|Dusseldorf|500
3|Moscow|1100
4|Sydney|200
5|Chennai|400

Prerequisites

• This tutorial runs Confluent Platform in Docker. Before proceeding: - Install Docker Desktop or Docker Engine (version 19.03.0 or later) if you don’t already have it - Install the Docker Compose plugin if you don’t already have it. This isn’t necessary if you have Docker Desktop since it includes Docker Compose. - Start Docker if it’s not already running, either by starting Docker Desktop or, if you manage Docker Engine with systemd, via systemctl - Verify that Docker is set up properly by ensuring no errors are output when you run docker info and docker compose version on the command line
• Maven command mvn to compile Java code
• timeout: used by the bash scripts to terminate a consumer process after a certain period of time. timeout is available on most Linux distributions but not on macOS. macOS users can install timeout via brew install coreutils.

Run example

1. Clone the examples GitHub repository and check out the 7.4.8-post branch.

   git clone https://github.com/confluentinc/examples
   cd examples
   git checkout 7.4.8-post
   

2. Change directory to the connect-streams-pipeline example.

   cd connect-streams-pipeline
   

3. Run the examples end-to-end:

   ./start.sh
   

4. Open your browser and navigate to the Confluent Control Center web interface Management -> Connect tab to see the data in the Kafka topics and the deployed connectors.

Example 1: Kafka console producer -> Key:String and Value:String

• Command line kafka-console-producer produces String keys and String values to a Kafka topic.
• Client application reads from the Kafka topic using Serdes.String() for both key and value.

Example 2: JDBC source connector with Single Message Transformations -> Key:Long and Value:JSON

• Kafka Connect JDBC source connector produces JSON values, and inserts the key using single message transformations, also known as SMTs. This is helpful because by default JDBC source connector does not insert a key.
• This example uses a few SMTs including one to cast the key to an int64. The key uses the org.apache.kafka.connect.converters.LongConverter provided by KAFKA-6913.

{
  "name": "test-source-sqlite-jdbc-autoincrement-jdbcjson",
  "config": {
    "connector.class": "io.confluent.connect.jdbc.JdbcSourceConnector",
    "tasks.max": "1",
    "key.converter": "org.apache.kafka.connect.converters.LongConverter",
    "key.converter.schemas.enable": "false",
    "value.converter": "org.apache.kafka.connect.json.JsonConverter",
    "value.converter.schemas.enable": "false",
    "transforms": "InsertKey, ExtractId, CastLong",
    "transforms.InsertKey.type": "org.apache.kafka.connect.transforms.ValueToKey",
    "transforms.InsertKey.fields": "id",
    "transforms.ExtractId.type": "org.apache.kafka.connect.transforms.ExtractField$Key",
    "transforms.ExtractId.field": "id",
    "transforms.CastLong.type": "org.apache.kafka.connect.transforms.Cast$Key",
    "transforms.CastLong.spec": "int64",

    "connection.url": "jdbc:sqlite:/usr/local/lib/retail.db",
    "mode": "incrementing",
    "incrementing.column.name": "id",
    "topic.prefix": "jdbcjson-",
    "table.whitelist": "locations"
  }
}

• Client application reads from the Kafka topic using Serdes.Long() for key and a custom JSON Serde for the value.

Example 3: JDBC source connector with SpecificAvro -> Key:String(null) and Value:SpecificAvro

• Kafka Connect JDBC source connector produces Avro values, and null String keys, to a Kafka topic.
• This example uses a single message transformation (SMT) called SetSchemaMetadata with code that has a fix for KAFKA-5164, allowing the connector to set the namespace in the schema. If you do not have the fix for KAFKA-5164, see Example 4 that uses GenericAvro instead of SpecificAvro.

{
  "name": "test-source-sqlite-jdbc-autoincrement-jdbcspecificavro",
  "config": {
    "connector.class": "io.confluent.connect.jdbc.JdbcSourceConnector",
    "tasks.max": "1",
    "key.converter": "org.apache.kafka.connect.converters.LongConverter",
    "key.converter.schemas.enable": "false",
    "value.converter": "io.confluent.connect.avro.AvroConverter",
    "value.converter.schema.registry.url": "http://schema-registry:8081",
    "value.converter.schemas.enable": "true",
    "transforms": "SetValueSchema",
    "transforms.SetValueSchema.type": "org.apache.kafka.connect.transforms.SetSchemaMetadata$Value",
    "transforms.SetValueSchema.schema.name": "io.confluent.examples.connectandstreams.avro.Location",

    "connection.url": "jdbc:sqlite:/usr/local/lib/retail.db",
    "mode": "incrementing",
    "incrementing.column.name": "id",
    "topic.prefix": "jdbcspecificavro-",
    "table.whitelist": "locations"
  }
}

• Client application reads from the Kafka topic using SpecificAvroSerde for the value and then the map function to convert the stream of messages to have Long keys and custom class values.

Example 4: JDBC source connector with GenericAvro -> Key:String(null) and Value:GenericAvro

• Kafka Connect JDBC source connector produces Avro values, and null String keys, to a Kafka topic.

{
  "name": "test-source-sqlite-jdbc-autoincrement-jdbcgenericavro",
  "config": {
    "connector.class": "io.confluent.connect.jdbc.JdbcSourceConnector",
    "tasks.max": "1",
    "key.converter": "org.apache.kafka.connect.converters.LongConverter",
    "key.converter.schemas.enable": "false",
    "value.converter": "io.confluent.connect.avro.AvroConverter",
    "value.converter.schema.registry.url": "http://schema-registry:8081",
    "value.converter.schemas.enable": "true",

    "connection.url": "jdbc:sqlite:/usr/local/lib/retail.db",
    "mode": "incrementing",
    "incrementing.column.name": "id",
    "topic.prefix": "jdbcgenericavro-",
    "table.whitelist": "locations"
  }
}

• Client application reads from the Kafka topic using GenericAvroSerde for the value and then the map function to convert the stream of messages to have Long keys and custom class values.
• This example currently uses GenericAvroSerde and not SpecificAvroSerde for a specific reason. JDBC source connector currently doesn’t set a namespace when it generates a schema name for the data it is producing to Kafka. For SpecificAvroSerde, the lack of namespace is a problem when trying to match reader and writer schema because Avro uses the writer schema name and namespace to create a classname and tries to load this class, but without a namespace, the class will not be found.

Example 5: Java client producer with SpecificAvro -> Key:Long and Value:SpecificAvro

• Java client produces Long keys and SpecificAvro values to a Kafka topic.
• Client application reads from the Kafka topic using Serdes.Long() for key and SpecificAvroSerde for the value.

Example 6: JDBC source connector with Avro to ksqlDB -> Key:Long and Value:Avro

• Kafka Connect JDBC source connector produces Avro values, and null keys, to a Kafka topic.

{
  "name": "test-source-sqlite-jdbc-autoincrement-jdbcavroksql",
  "config": {
    "connector.class": "io.confluent.connect.jdbc.JdbcSourceConnector",
    "tasks.max": "1",
    "key.converter": "org.apache.kafka.connect.json.JsonConverter",
    "key.converter.schemas.enable": "false",
    "value.converter": "io.confluent.connect.avro.AvroConverter",
    "value.converter.schema.registry.url": "http://schema-registry:8081",
    "value.converter.schemas.enable": "true",

    "connection.url": "jdbc:sqlite:/usr/local/lib/retail.db",
    "mode": "incrementing",
    "incrementing.column.name": "id",
    "topic.prefix": "jdbcavroksql-",
    "table.whitelist": "locations"
  }
}

• ksqlDB reads from the Kafka topic and then uses PARTITION BY to create a new stream of messages with BIGINT keys.

Technical Notes

• KAFKA-5245: one needs to provide the Serdes twice, (1) when calling StreamsBuilder#stream() and (2) when calling KStream#groupByKey()
• PR-531: Confluent distribution provides packages for GenericAvroSerde and SpecificAvroSerde
• KAFKA-2378: adds APIs to be able to embed Kafka Connect into client applications
• KAFKA-2526: one cannot use the --key-serializer argument in kafka-console-producer to serialize the key as a Long. As a result, in this example the key is serialized as a String. As a workaround, you could write your own kafka.common.MessageReader (e.g. check out the default implementation of LineMessageReader) and then you can specify --line-reader argument in kafka-console-producer.
• KAFKA-5164: allows the connector to set the namespace in the schema.