Node.js bindings for librdkafka with Avro schema serialization.

The kafka-avro library is a wrapper that combines the node-rdkafka and avsc libraries to allow for Production and Consumption of messages on kafka validated and serialized by Avro.

Install the module using NPM:

npm install kafka-avro --save

The kafka-avro library operates in the following steps:

1. You provide your Kafka Brokers and Schema Registry (SR) Url to a new instance of kafka-avro.
2. You initialize kafka-avro, that will tell the library to query the SR for all registered schemas, evaluate and store them in runtime memory.
3. kafka-avro will then expose the getConsumer() and getProducer() methods, which both return instances of the corresponding Constructors from the node-rdkafka library.

The instances of "node-rdkafka" that are returned by kafka-avro are hacked so as to intercept produced and consumed messages and run them by the Avro de/serializer along with Confluent's Schema Registry Magic Byte and Schema Id.

You are highly encouraged to read the "node-rdkafka" documentation, as it explains how you can use the Producer and Consumer instances as well as check out the available configuration options of node-rdkafka.

The Kafka.CODES enumeration of constant values provided by the "node-rdkafka" library is also available as a static var at:

var KafkaAvro = require('kafka-avro');

console.log(KafkaAvro.CODES);

var KafkaAvro = require('kafka-avro');

var kafkaAvro = new KafkaAvro({
    kafkaBroker: 'localhost:9092',
    schemaRegistry: 'https://localhost:8081',
});

// Query the Schema Registry for all topic-schema's
// fetch them and evaluate them.
kafkaAvro.init()
    .then(function() {
        console.log('Ready to use');
    });

When instantiating kafka-avro you may pass the following options:

• kafkaBroker String REQUIRED The url or comma delimited strings pointing to your kafka brokers.
• schemaRegistry String REQUIRED The url to the Schema Registry.
• topics Array of Strings You may optionally define specific topics to be fetched by kafka-avro vs fetching schemas for all the topics which is the default behavior.
• fetchAllVersions Boolean Set to true to fetch all versions for each topic, use it when updating of schemas is often in your environment.
• fetchRefreshRate Number The pooling time (in seconds) to the schemas be fetched and updated in background. This is useful to keep with schemas changes in production. The default value is 0 seconds (disabled).
• parseOptions Object Schema parse options to pass to avro.parse(). parseOptions.wrapUnions is set to true by default.
• httpsAgent Object initialized https Agent class

NOTICE: You need to initialize kafka-avro before you can produce or consume messages.

By invoking the kafkaAvro.getProducer() method, kafka-avro will instantiate a Producer, make it connect and wait for it to be ready before the promise is resolved.

kafkaAvro.getProducer({
  // Options listed bellow
})
    // "getProducer()" returns a Bluebird Promise.
    .then(function(producer) {
        var topicName = 'test';

        producer.on('disconnected', function(arg) {
          console.log('producer disconnected. ' + JSON.stringify(arg));
        });

        var value = {name:'John'};
        var key = 'key';

        // if partition is set to -1, librdkafka will use the default partitioner
        var partition = -1;
        producer.produce(topicName, partition, value, key);
    })

What kafka-avro basically does is wrap around node-rdkafka and intercept the produce method to validate and serialize the message.

• node-rdkafka Producer Tutorial
• Full list of Producer's options

NOTICE: You need to initialize kafka-avro before you can produce or consume messages.

By invoking the kafkaAvro.getConsumer() method, kafka-avro will instantiate a Consumer, listen on log, error and disconnect events and return it to you. Depending on the consuming pattern you follow you may or may not need to perform a connect().

When consuming topics using the data event you will need to perform a connect() as per node-rdkafka documentation:

kafkaAvro.getConsumer({
  'group.id': 'librd-test',
  'socket.keepalive.enable': true,
  'enable.auto.commit': true,
})
  // the "getConsumer()" method will return a bluebird promise.
  .then(function(consumer) {
    // Perform a consumer.connect()
    return new Promise(function (resolve, reject) {
      consumer.on('ready', function() {
        resolve(consumer);
      });

      consumer.connect({}, function(err) {
        if (err) {
          reject(err);
          return;
        }
        resolve(consumer); // depend on Promises' single resolve contract.
      });
    });
  })
  .then(function(consumer) {
    // Subscribe and consume.
    var topicName = 'test';
    consumer.subscribe([topicName]);
    consumer.consume();
    consumer.on('data', function(rawData) {
      console.log('data:', rawData);
    });
  });

kafkaAvro.getConsumerStream({
  'group.id': 'librd-test',
  'socket.keepalive.enable': true,
  'enable.auto.commit': true,
},
{
  'request.required.acks': 1
},
{
  'topics': 'test'
})
  .then(function(stream) {
      stream.on('error', function(err) {
        if (err) console.log(err);
        process.exit(1);
      });

      stream.on('data', function (rawData) {
          console.log('data:', rawData)
      });

      stream.on('error', function(err) {
        console.log(err);
        process.exit(1);
      });

      stream.consumer.on('event.error', function(err) {
        console.log(err);
      })
  });

Same deal here, thin wrapper around node-rdkafka and deserialize incoming messages before they reach your consuming method.

• node-rdkafka Consumer Tutorial
• Full list of Consumer's options

kafka-avro intercepts all incoming messages and augments the object with one more property named parsed which contained the avro deserialized object. Here is a breakdown of the properties included in the message object you receive when consuming messages:

• value Buffer The raw message buffer from Kafka.
• size Number The size of the message.
• key String|Number Partioning key used.
• topic String The topic this message comes from.
• offset Number The Kafka offset.
• partition Number The kafka partion used.
• parsed Object The avro deserialized message as a JS Object ltieral.
• schemaId Number The Registry Schema id of the consumed message.

The KafkaAvro instance also provides the following methods:

The Kafka Avro library logs messages using the Bunyan logger. To enable logging you will have to define at least one of the needed ENV variables:

• KAFKA_AVRO_LOG_LEVEL Set it a valid Bunyan log level value to activate console logging (Typically you'd need either info or debug as values.)
• KAFKA_AVRO_LOG_NO_COLORS Set this to any value to disable color when logging.

WARNING The logger will not emit any messages as it was expected, there is an open issue on Bunyan's repository pending a solution on this. So no logging for now.

NOTICE This is a static method on the KafkaAvro constructor, not the instance. Therefore there is a single logger instance for the whole runtime.

Returns {Bunyan.Logger} Bunyan logger instance.

var KafkaAvro = require('kafka-avro');
var fmt = require('bunyan-format');

var kafkaLog  = KafkaAvro.getLogger();

kafkaLog.addStream({
    type: 'stream',
    stream: fmt({
        outputMode: 'short',
        levelInString: true,
    }),
    level: 'info',
});

Read more about the bunyan-format package.

Serialize the provided value with Avro, including the magic Byte and schema id provided.

Returns {Buffer} Serialized buffer message.

• type {avsc.Type} The avro type instance.
• schemaId {number} The Schema Id in the SR.
• value {*} Any value to serialize.

Deserialize the provided message, expects a message that includes Magic Byte and schema id.

Returns {*} Deserialized message.

• type {avsc.Type} The avro type instance.
• message {Buffer} Message in byte code.

You can use docker-compose up to up all the stack before you call your integration tests with npm test. How the integration tests are outside the containers, you will need set you hosts file to :

127.0.0.1 kafka

1. Update the changelog bellow.
2. Ensure you are on master.
3. Type: grunt release
   • grunt release:minor for minor number jump.
   • grunt release:major for major number jump.

• 1.1.3, 02 January 2019
  • Fixing broken fetchAllVersions feature (by ricardohbin)
• 1.1.2, 08 November 2018
  • Handle topics which use identical schemas (by scottwd9)
• 1.1.1, 23 August 2018
  • Set schemaMeta for key schemas also (by eparreno)
• 1.1.0, 06 August 2018
  • Updating to node-rdkafka v2.3.4 (by ricardohbin)
• 1.0.6, 11 July 2018
  • Adding the fetchRefreshRate parameter, to set a way to update the schemas after the app initialization. (by ricardohbin)
• 1.0.5, 27 June 2018
  • Fixes kafka-producer to pass timestamp and opaque correctly (by javierholguera)
• 1.0.4, 30 May 2018
  • Allowing schema-registry urls with paths (by 941design)
• 1.0.3, 28 May 2018
  • Added support for providing configured https.Agent to support schema-registry using private certificates. (by scottchapman)
• 1.0.2, 23 Apr 2018
  • Using URL module instead of strings to resolve schema registry urls. (by ricardohbin)
• 1.0.1, 03 Apr 2018
  • Breaking change: New consumer stream API, changes in producer and fixing all integration tests (by ricardohbin).
• 1.0.0, 28 Mar 2018
  • Updating docs and the libs avsc (v5.2.3) and node-rdkafka (v2.1.3) (by ricardohbin).
• v0.8.1, 01 Feb 2018
  • Allow customization of AVSC parse options (thank you dapetcu21).
• v0.8.0, 27 Nov 2017
  • Provides option to fetch all past versions of a topic (thank you CMTegner).
  • Provides option to select which topics should be fetched.
• v0.7.0, 24 Feb 2017
  • New mechanism in deserializing messages, the schema id will now be parsed from the message and if this schema id is found in the local registry kafka-avro will use that schema to deserialize. If it is not found then it will use the provided schema type, which would be the last known for the topic.
  • Added schemaId property on the consumed messages.
• v0.6.4, 23 Feb 2017
  • Catch errors thrown by the deserializer.
• v0.6.3, 20 Feb 2017
  • Will now pass topic options on getConsumer and getProducer methods.
• v0.6.2, 18 Feb 2017
  • Fixed Magic Byte encoding for large payloads.
• v0.6.1, 17 Feb 2017
  • Don't log consumer log messages (logger, not kafka).
• v0.6.0, 16 Feb 2017
  • Breaking change: Consumers will no longer auto-connect, you are required to perform the connect() method manually, check the docs.
  • Refactored logging, you can now enable it using an env var, check docs on Logging.
• v0.5.1, 15 Feb 2017
  • Catch errors from connect() callbacks for both Consumer and Producer.
• v0.5.0, 15 Feb 2017
  • Upgrade to node-rdkafka 0.7.0-ALPHA.3 which changes the consumer API by decoupling subscribing from consuming.
• v0.4.3, 15 Feb 2017
  • Locked this version to 0.7.0-ALPHA.2 of node-rdkafka which broke BC in 0.7.0-ALPHA.3.
• v0.4.2, 15 Feb 2017
  • Fixed connect() invocation for consumers and producers.
• v0.4.1, 10 Feb 2017
  • Fixed relaying Kafka consumer logs.
• v0.4.0, 03 Feb 2017
  • Refactored all logging to use a central Bunyan logger that is now provided through the static method KafkaAvro.getLogger().
  • Allowed for an Array of strings as topic argument for Consumer's getReadStream() method.
• v0.3.0, 01 Feb 2017
  • Now force uses Magic Byte in any occasion when de/serializing.
  • Exposed serialize() and deserialize() methods.
  • Fixed de/serializing of topics not found in the Schema Registry.
  • Tweaked log namespaces, still more work required to eventize them.
• v0.2.0, 30 Jan 2017
  • Added Confluent's Magic Byte support when encoding and decoding messages.
• v0.1.2, 27 Jan 2017
  • Suppress schema parsing errors.
• v0.1.1, 27 Jan 2017
  • Fix signature of getConsumer() method.
• v0.1.1, 27 Jan 2017
  • Expose CODES enum from node-rdkafka.
  • Write more docs, add the event based consuming method.
• v0.1.0, 26 Jan 2017
  • First fully working release.
• v0.0.1, 25 Jan 2017
  • Big Bang

Copyright Waldo, Inc. Licensed under the MIT.