How do I create the json for creating a distributed Kafka Connect Instance with a transformation?

Question

Using standalone mode I create a connector and my customized transformation like that:

name=rabbitmq-source
connector.class=com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector
tasks.max=1
rabbitmq.host=rabbitmq-server
rabbitmq.queue=answers
kafka.topic=net.gutefrage.answers
transforms=extractFields
transforms.extractFields.type=net.gutefrage.connector.transforms.ExtractFields$Value
transforms.extractFields.fields=body,envelope.routingKey
transforms.extractFields.structName=net.gutefrage.events

But for a distributed connector what is the syntax for the PUT request to the Connect REST API? I cannot find any example in the docs.

Already tried a couple of things like:

cat <<EOF >/tmp/connector
{
  "name": "rabbitmq-source",
  "config": {
    "connector.class": "com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector",
    "tasks.max": "1",
    "rabbitmq.host": "rabbitmq-server",
    "rabbitmq.queue": "answers",
    "kafka.topic": "net.gutefrage.answers",
    "transforms": "extractFields",
    "transforms.extractFields": {
      "type": "net.gutefrage.connector.transforms.ExtractFields$Value",
      "fields": "body,envelope.routingKey",
      "structName": "net.gutefrage.events"
    }
  }
}
EOF

curl -vs --stderr - -X POST -H "Content-Type: application/json" --data @/tmp/connector "http://localhost:8083/connectors"
rm /tmp/connector

or also this did not work:

{
  "name": "rabbitmq-source",
  "config": {
    "connector.class": "com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector",
    "tasks.max": "1",
    "rabbitmq.host": "rabbitmq-server",
    "rabbitmq.queue": "answers",
    "kafka.topic": "net.gutefrage.answers",
    "transforms": "extractFields",
    "transforms.extractFields.type": "net.gutefrage.connector.transforms.ExtractFields$Value",
    "transforms.extractFields.fields": "body,envelope.routingKey",
    "transforms.extractFields.structName": "net.gutefrage.events"
  }
}

For the last variant I get the following error:

{"error_code":400,"message":"Connector configuration is invalid and contains the following 1 error(s):\nInvalid value class net.gutefrage.connector.transforms.ExtractFields for configuration transforms.extractFields.type: Error getting config definition from Transformation: null\nYou can also find the above list of errors at the endpoint `/{connectorType}/config/validate`"}

Please note that with the properties format it works nicely (using Landoops Create New Connector UI in fast-data-dev. Interesting that the Landoop's Ui feature 'translate to curl' produces the very same json as my second example)

Update

To be sure that it's not a problem with Landoop, docker and with my custom transformation, I've started zookeeper, broker, schema registry and Kafka Connect in distributed mode with the standard distributed properties from COP 3.3.0

bin/connect-distributed etc/schema-registry/connect-avro-distributed.properties

which logs [2017-09-13 14:07:52,930] INFO Loading plugin from: /opt/connectors/confluent-oss-gf-assembly-1.0.jar (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader:176) [2017-09-13 14:07:53,711] INFO Registered loader: PluginClassLoader{pluginLocation=file:/opt/connectors/confluent-oss-gf-assembly-1.0.jar} (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader:199) [2017-09-13 14:07:53,711] INFO Added plugin 'com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector' (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader:132) [2017-09-13 14:07:53,712] INFO Added plugin 'net.gutefrage.connector.transforms.ExtractFields$Key' (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader:132) [2017-09-13 14:07:53,712] INFO Added plugin 'net.gutefrage.connector.transforms.ExtractFields$Value' (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader:132) All good so far. Then I created a connector config:

cat <<EOF >/tmp/connector
{ "name": "rabbitmq-source", "config": { "connector.class": "com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector", "tasks.max": "1", "rabbitmq.host": "rabbitmq-server", "rabbitmq.queue": "answers", "kafka.topic": "net.gutefrage.answers", "transforms": "extractFields", "transforms.extractFields.type": "org.apache.kafka.connect.transforms.ExtractField$Value", "transforms.extractFields.field": "body" } } EOF Please note that there I do now use the standard (bundled) extract field transform. When I post that with curl -vs --stderr - -X POST -H "Content-Type: application/json" --data @/tmp/connector "http://localhost:8083/connectors" I get the same

{"error_code":400,"message":"Connector configuration is invalid and contains the following 1 error(s):\nInvalid value class org.apache.kafka.connect.transforms.ExtractField for configuration transforms.extractFields.type: Error getting config definition from Transformation: null\nYou can also find the above list of errors at the endpoint `/{connectorType}/config/validate`"}*

Answer 1

Make sure the $Value in transforms.extractFields.type=net.gutefrage.connector.transforms.ExtractFields$Value is not interpreted as a variable by the bash command cat. It worked for me.

Answer 2

With fast-data-dev you can build a JAR file for any connector, and then just add it in the classpath with the instructions at

https://github.com/Landoop/fast-data-dev#enable-additional-connectors

The UI will auto-detect the new connector - and provide you instructions when you hit NEW for the new connector at:

http://localhost:3030/kafka-connect-ui

What would also be worth trying - as fast-data-dev comes already with an generic MQTT sink connector, is trying it out. See instructions at http://docs.datamountaineer.com/en/latest/mqtt-sink.html

You would effectively need to do connect.mqtt.kcql=INSERT INTO /answers SELECT body FROM net.gutefrage.answers

As this is a generic MQTT connector - possibly you will need to add the rabbitmq client library using the enable-additional-connectors instructions

Answer 3

For using json format of connector config and the CP connect CLI, the jq tool has to be installed on the machine where the Kafka-Connect Cluster is running.

E.g. for Landoops fast-data-dev environment you'll have to

docker exec rabbitmqconnect_fast-data-dev_1 apk add --no-cache jq

Then this will work:

docker exec rabbitmqconnect_fast-data-dev_1 /opt/confluent-3.3.0/bin/confluent config rabbitmq-source -d /tmp/connector-config.json

This doesn't solve the issue when using the connector REST endpoint though.

Answer 4

If you want to run the Kafka Connect worker in standalone mode, then you must start the worker and supply the worker configuration file and one or more connector configuration files. All of those configuration files are in Java properties format, so the first configuration sample you provided is the correct format:

name=rabbitmq-source
connect.class=com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector
tasks.max=1
rabbitmq.host=rabbitmq-server
rabbitmq.queue=answers
kafka.topic=net.gutefrage.answers
transforms=extractFields
transforms.extractFields.type=net.gutefrage.connector.transforms.ExtractFields$Value
transforms.extractFields.fields=body,envelope.routingKey
transforms.extractFields.structName=net.gutefrage.events

If you want to run the Kafka Connect worker in distributed mode, then you will have to first start the distributed worker and then create the connector as a second step using the REST API and a PUT request with a JSON document to the /connectors endpoint. That JSON document would match the format of you're second JSON document:

{
  "name": "rabbitmq-source",
  "config": {
    "connector.class": "com.github.jcustenborder.kafka.connect.rabbitmq.RabbitMQSourceConnector",
    "tasks.max": "1",
    "rabbitmq.host": "rabbitmq-server",
    "rabbitmq.queue": "answers",
    "kafka.topic": "net.gutefrage.answers",
    "transforms": "extractFields",
    "transforms.extractFields.type": "net.gutefrage.connector.transforms.ExtractFields$Value",
    "transforms.extractFields.fields": "body,envelope.routingKey",
    "transforms.extractFields.structName": "net.gutefrage.events"
  }
}

The Confluent CLI, included in Confluent's Open Source Platform that includes Kafka, is a developer tool to help you quickly get started by running a Zookeeper instance, a Kafka broker, the Confluent Schema Registry, the REST proxy, and a Connect worker in distributed mode. When you load a connector, you specify the connector configuration as either a JSON file or property file, converting the latter to a JSON format using jq.

However, the error you reported is:

{
  "error_code":400,
  "message":"Connector configuration is invalid and contains the following 1 error(s):\nInvalid value class net.gutefrage.connector.transforms.ExtractFields for configuration transforms.extractFields.type: Error getting config definition from Transformation: null\nYou can also find the above list of errors at the endpoint `/{connectorType}/config/validate`"
}

The important part of this error message is "Error getting config definition from Transformation: null". Although this is a bit too cryptic, it means that the config() method of the net.gutefrage.connector.transforms.ExtractFields Java class is returning null.

Make sure that the net.gutefrage.connector.transforms.ExtractFields$Valuestring you specified is the correct fully qualified name for the nested static class Value, and that the Value class fully and correctly implements the org.apache.kafka.connect.transforms.Transformation<? extends ConnectRecord<R>> interface. Note that the config() method must return a non-null ConfigDef object.

Take a look at this example of a Single Message Transform (SMT) that ships with Apache Kafka, or Robin's blog post for other examples.