node-rdkafka - Node.js wrapper for Kafka C/C++ library

Copyright (c) 2016 Blizzard Entertainment.

https://github.com/blizzard/node-rdkafka

Looking for Collaborators!

I am looking for your help to make this project even better! If you're interested, check this out

Overview

The node-rdkafka library is a high-performance NodeJS client for Apache Kafka that wraps the native librdkafka library. All the complexity of balancing writes across partitions and managing (possibly ever-changing) brokers should be encapsulated in the library.

This library currently uses librdkafka version 2.1.1.

Reference Docs

To view the reference docs for the current version, go here

Contributing

For guidelines on contributing please see CONTRIBUTING.md

Code of Conduct

Play nice; Play fair.

Requirements

• Apache Kafka >=0.9
• Node.js >=4
• Linux/Mac
• Windows?! See below
• OpenSSL

Mac OS High Sierra / Mojave

OpenSSL has been upgraded in High Sierra and homebrew does not overwrite default system libraries. That means when building node-rdkafka, because you are using openssl, you need to tell the linker where to find it:

export CPPFLAGS=-I/usr/local/opt/openssl/include
export LDFLAGS=-L/usr/local/opt/openssl/lib

Then you can run npm install on your application to get it to build correctly.

NOTE: From the librdkafka docs

WARNING: Due to a bug in Apache Kafka 0.9.0.x, the ApiVersionRequest (as sent by the client when connecting to the broker) will be silently ignored by the broker causing the request to time out after 10 seconds. This causes client-broker connections to stall for 10 seconds during connection-setup before librdkafka falls back on the broker.version.fallback protocol features. The workaround is to explicitly configure api.version.request to false on clients communicating with <=0.9.0.x brokers.

Alpine

Using Alpine Linux? Check out the docs.

Windows

Windows build is not compiled from librdkafka source but it is rather linked against the appropriate version of NuGet librdkafka.redist static binary that gets downloaded from https://globalcdn.nuget.org/packages/librdkafka.redist.2.1.1.nupkg during installation. This download link can be changed using the environment variable NODE_RDKAFKA_NUGET_BASE_URL that defaults to https://globalcdn.nuget.org/packages/ when it's no set.

Requirements:

• node-gyp for Windows

Note: I still do not recommend using node-rdkafka in production on Windows. This feature was in high demand and is provided to help develop, but we do not test against Windows, and windows support may lag behind Linux/Mac support because those platforms are the ones used to develop this library. Contributors are welcome if any Windows issues are found :)

Tests

This project includes two types of unit tests in this project:

• end-to-end integration tests
• unit tests

You can run both types of tests by using Makefile. Doing so calls mocha in your locally installed node_modules directory.

• Before you run the tests, be sure to init and update the submodules:
  1. git submodule init
  2. git submodule update
• To run the unit tests, you can run make lint or make test.
• To run the integration tests, you must have a running Kafka installation available. By default, the test tries to connect to localhost:9092; however, you can supply the KAFKA_HOST environment variable to override this default behavior. Run make e2e.

Usage

You can install the node-rdkafka module like any other module:

npm install node-rdkafka

To use the module, you must require it.

var Kafka = require('node-rdkafka');

Configuration

You can pass many configuration options to librdkafka. A full list can be found in librdkafka's Configuration.md

Configuration keys that have the suffix _cb are designated as callbacks. Some of these keys are informational and you can choose to opt-in (for example, dr_cb). Others are callbacks designed to return a value, such as partitioner_cb.

Not all of these options are supported. The library will throw an error if the value you send in is invalid.

The library currently supports the following callbacks:

• partitioner_cb
• dr_cb or dr_msg_cb
• event_cb
• rebalance_cb (see Rebalancing)
• offset_commit_cb (see Commits)

Librdkafka Methods

This library includes two utility functions for detecting the status of your installation. Please try to include these when making issue reports where applicable.

You can get the features supported by your compile of librdkafka by reading the variable "features" on the root of the node-rdkafka object.

const Kafka = require('node-rdkafka');
console.log(Kafka.features);

// #=> [ 'gzip', 'snappy', 'ssl', 'sasl', 'regex', 'lz4' ]

You can also get the version of librdkafka

const Kafka = require('node-rdkafka');
console.log(Kafka.librdkafkaVersion);

// #=> 2.1.1

Sending Messages

A Producer sends messages to Kafka. The Producer constructor takes a configuration object, as shown in the following example:

var producer = new Kafka.Producer({
  'metadata.broker.list': 'kafka-host1:9092,kafka-host2:9092'
});

A Producer requires only metadata.broker.list (the Kafka brokers) to be created. The values in this list are separated by commas. For other configuration options, see the Configuration.md file described previously.

The following example illustrates a list with several librdkafka options set.

var producer = new Kafka.Producer({
  'client.id': 'kafka',
  'metadata.broker.list': 'localhost:9092',
  'compression.codec': 'gzip',
  'retry.backoff.ms': 200,
  'message.send.max.retries': 10,
  'socket.keepalive.enable': true,
  'queue.buffering.max.messages': 100000,
  'queue.buffering.max.ms': 1000,
  'batch.num.messages': 1000000,
  'dr_cb': true
});

Stream API

You can easily use the Producer as a writable stream immediately after creation (as shown in the following example):

// Our producer with its Kafka brokers
// This call returns a new writable stream to our topic 'topic-name'
var stream = Kafka.Producer.createWriteStream({
  'metadata.broker.list': 'kafka-host1:9092,kafka-host2:9092'
}, {}, {
  topic: 'topic-name'
});

// Writes a message to the stream
var queuedSuccess = stream.write(Buffer.from('Awesome message'));

if (queuedSuccess) {
  console.log('We queued our message!');
} else {
  // Note that this only tells us if the stream's queue is full,
  // it does NOT tell us if the message got to Kafka!  See below...
  console.log('Too many messages in our queue already');
}

// NOTE: MAKE SURE TO LISTEN TO THIS IF YOU WANT THE STREAM TO BE DURABLE
// Otherwise, any error will bubble up as an uncaught exception.
stream.on('error', function (err) {
  // Here's where we'll know if something went wrong sending to Kafka
  console.error('Error in our kafka stream');
  console.error(err);
})

If you do not want your code to crash when an error happens, ensure you have an error listener on the stream. Most errors are not necessarily fatal, but the ones that are will immediately destroy the stream. If you use autoClose, the stream will close itself at the first sign of a problem.

Standard API

The Standard API is more performant, particularly when handling high volumes of messages. However, it requires more manual setup to use. The following example illustrates its use:

var producer = new Kafka.Producer({
  'metadata.broker.list': 'localhost:9092',
  'dr_cb': true
});

// Connect to the broker manually
producer.connect();

// Wait for the ready event before proceeding
producer.on('ready', function() {
  try {
    producer.produce(
      // Topic to send the message to
      'topic',
      // optionally we can manually specify a partition for the message
      // this defaults to -1 - which will use librdkafka's default partitioner (consistent random for keyed messages, random for unkeyed messages)
      null,
      // Message to send. Must be a buffer
      Buffer.from('Awesome message'),
      // for keyed messages, we also specify the key - note that this field is optional
      'Stormwind',
      // you can send a timestamp here. If your broker version supports it,
      // it will get added. Otherwise, we default to 0
      Date.now(),
      // you can send an opaque token here, which gets passed along
      // to your delivery reports
    );
  } catch (err) {
    console.error('A problem occurred when sending our message');
    console.error(err);
  }
});

// Any errors we encounter, including connection errors
producer.on('event.error', function(err) {
  console.error('Error from producer');
  console.error(err);
})

// We must either call .poll() manually after sending messages
// or set the producer to poll on an interval (.setPollInterval).
// Without this, we do not get delivery events and the queue
// will eventually fill up.
producer.setPollInterval(100);

To see the configuration options available to you, see the Configuration section.

Methods

Events

Some configuration properties that end in _cb indicate that an event should be generated for that option. You can either:

• provide a value of true and react to the event
• provide a callback function directly

The following example illustrates an event:

var producer = new Kafka.Producer({
  'client.id': 'my-client', // Specifies an identifier to use to help trace activity in Kafka
  'metadata.broker.list': 'localhost:9092', // Connect to a Kafka instance on localhost
  'dr_cb': true // Specifies that we want a delivery-report event to be generated
});

// Poll for events every 100 ms
producer.setPollInterval(100);

producer.on('delivery-report', function(err, report) {
  // Report of delivery statistics here:
  //
  console.log(report);
});

The following table describes types of events.

Higher Level Producer

The higher level producer is a variant of the producer which can propagate callbacks to you upon message delivery.

var producer = new Kafka.HighLevelProducer({
  'metadata.broker.list': 'localhost:9092',
});

This will enrich the produce call so it will have a callback to tell you when the message has been delivered. You lose the ability to specify opaque tokens.

producer.produce(topicName, null, Buffer.from('alliance4ever'), null, Date.now(), (err, offset) => {
  // The offset if our acknowledgement level allows us to receive delivery offsets
  console.log(offset);
});

Additionally you can add serializers to modify the value of a produce for a key or value before it is sent over to Kafka.

producer.setValueSerializer(function(value) {
  return Buffer.from(JSON.stringify(value));
});

Otherwise the behavior of the class should be exactly the same.

Kafka.KafkaConsumer

To read messages from Kafka, you use a KafkaConsumer. You instantiate a KafkaConsumer object as follows:

var consumer = new Kafka.KafkaConsumer({
  'group.id': 'kafka',
  'metadata.broker.list': 'localhost:9092',
}, {});

The first parameter is the global config, while the second parameter is the topic config that gets applied to all subscribed topics. To view a list of all supported configuration properties, see the Configuration.md file described previously. Look for the C and * keys.

The group.id and metadata.broker.list properties are required for a consumer.

Rebalancing

Rebalancing is managed internally by librdkafka by default. If you would like to override this functionality, you may provide your own logic as a rebalance callback.

var consumer = new Kafka.KafkaConsumer({
  'group.id': 'kafka',
  'metadata.broker.list': 'localhost:9092',
  'rebalance_cb': function(err, assignment) {

    if (err.code === Kafka.CODES.ERRORS.ERR__ASSIGN_PARTITIONS) {
      // Note: this can throw when you are disconnected. Take care and wrap it in
      // a try catch if that matters to you
      this.assign(assignment);
    } else if (err.code == Kafka.CODES.ERRORS.ERR__REVOKE_PARTITIONS){
      // Same as above
      this.unassign();
    } else {
      // We had a real error
      console.error(err);
    }

  }
})

this is bound to the KafkaConsumer you have created. By specifying a rebalance_cb you can also listen to the rebalance event as an emitted event. This event is not emitted when using the internal librdkafka rebalancer.

Commits

When you commit in node-rdkafka, the standard way is to queue the commit request up with the next librdkafka request to the broker. When doing this, there isn't a way to know the result of the commit. Luckily there is another callback you can listen to to get this information

var consumer = new Kafka.KafkaConsumer({
  'group.id': 'kafka',
  'metadata.broker.list': 'localhost:9092',
  'offset_commit_cb': function(err, topicPartitions) {

    if (err) {
      // There was an error committing
      console.error(err);
    } else {
      // Commit went through. Let's log the topic partitions
      console.log(topicPartitions);
    }

  }
})

this is bound to the KafkaConsumer you have created. By specifying an offset_commit_cb you can also listen to the offset.commit event as an emitted event. It receives an error and the list of topic partitions as argument. This is not emitted unless opted in.

Message Structure

Messages that are returned by the KafkaConsumer have the following structure.

{
  value: Buffer.from('hi'), // message contents as a Buffer
  size: 2, // size of the message, in bytes
  topic: 'librdtesting-01', // topic the message comes from
  offset: 1337, // offset the message was read from
  partition: 1, // partition the message was on
  key: 'someKey', // key of the message if present
  timestamp: 1510325354780 // timestamp of message creation
}

Stream API

The stream API is the easiest way to consume messages. The following example illustrates the use of the stream API:

// Read from the librdtesting-01 topic... note that this creates a new stream on each call!
var stream = KafkaConsumer.createReadStream(globalConfig, topicConfig, {
  topics: ['librdtesting-01']
});

stream.on('data', function(message) {
  console.log('Got message');
  console.log(message.value.toString());
});

You can also get the consumer from the streamConsumer, for using consumer methods. The following example illustrates that:

stream.consumer.commit(); // Commits all locally stored offsets

Standard API

You can also use the Standard API and manage callbacks and events yourself. You can choose different modes for consuming messages:

• Flowing mode. This mode flows all of the messages it can read by maintaining an infinite loop in the event loop. It only stops when it detects the consumer has issued the unsubscribe or disconnect method.
• Non-flowing mode. This mode reads a single message from Kafka at a time manually.

The following example illustrates flowing mode:

// Flowing mode
consumer.connect();

consumer
  .on('ready', function() {
    consumer.subscribe(['librdtesting-01']);

    // Consume from the librdtesting-01 topic. This is what determines
    // the mode we are running in. By not specifying a callback (or specifying
    // only a callback) we get messages as soon as they are available.
    consumer.consume();
  })
  .on('data', function(data) {
    // Output the actual message contents
    console.log(data.value.toString());
  });

The following example illustrates non-flowing mode:

// Non-flowing mode
consumer.connect();

consumer
  .on('ready', function() {
    // Subscribe to the librdtesting-01 topic
    // This makes subsequent consumes read from that topic.
    consumer.subscribe(['librdtesting-01']);

    // Read one message every 1000 milliseconds
    setInterval(function() {
      consumer.consume(1);
    }, 1000);
  })
  .on('data', function(data) {
    console.log('Message found!  Contents below.');
    console.log(data.value.toString());
  });

The following table lists important methods for this API.

The following table lists events for this API.

Reading current offsets from the broker for a topic

Some times you find yourself in the situation where you need to know the latest (and earliest) offset for one of your topics. Connected producers and consumers both allow you to query for these through queryWaterMarkOffsets like follows:

var timeout = 5000, partition = 0;
consumer.queryWatermarkOffsets('my-topic', partition, timeout, function(err, offsets) {
  var high = offsets.highOffset;
  var low = offsets.lowOffset;
});

producer.queryWatermarkOffsets('my-topic', partition, timeout, function(err, offsets) {
  var high = offsets.highOffset;
  var low = offsets.lowOffset;
});

An error will be returned if the client was not connected or the request timed out within the specified interval.

Metadata

Both Kafka.Producer and Kafka.KafkaConsumer include a getMetadata method to retrieve metadata from Kafka.

Getting metadata on any connection returns the following data structure:

{
  orig_broker_id: 1,
  orig_broker_name: "broker_name",
  brokers: [
    {
      id: 1,
      host: 'localhost',
      port: 40
    }
  ],
  topics: [
    {
      name: 'awesome-topic',
      partitions: [
        {
          id: 1,
          leader: 20,
          replicas: [1, 2],
          isrs: [1, 2]
        }
      ]
    }
  ]
}

The following example illustrates how to use the getMetadata method.

When fetching metadata for a specific topic, if a topic reference does not exist, one is created using the default config. Please see the documentation on Client.getMetadata if you want to set configuration parameters, e.g. acks, on a topic to produce messages to.

var opts = {
  topic: 'librdtesting-01',
  timeout: 10000
};

producer.getMetadata(opts, function(err, metadata) {
  if (err) {
    console.error('Error getting metadata');
    console.error(err);
  } else {
    console.log('Got metadata');
    console.log(metadata);
  }
});

Admin Client

node-rdkafka now supports the admin client for creating, deleting, and scaling out topics. The librdkafka APIs also support altering configuration of topics and broker, but that is not currently implemented.

To create an Admin client, you can do as follows:

const Kafka = require('node-rdkafka');

const client = Kafka.AdminClient.create({
  'client.id': 'kafka-admin',
  'metadata.broker.list': 'broker01'
});

This will instantiate the AdminClient, which will allow the calling of the admin methods.

client.createTopic({
  topic: topicName,
  num_partitions: 1,
  replication_factor: 1
}, function(err) {
  // Done!
});

All of the admin api methods can have an optional timeout as their penultimate parameter.

The following table lists important methods for this API.

Check the tests for an example of how to use this API!