Rdkafka

The rdkafka gem is a modern Kafka client library for Ruby based on librdkafka. It wraps the production-ready C client using the ffi gem and targets Kafka 1.0+ and Ruby versions under security or active maintenance. We remove a Ruby version from our CI builds when they become EOL.

rdkafka was written because of the need for a reliable Ruby client for Kafka that supports modern Kafka at AppSignal. AppSignal runs it in production on very high-traffic systems.

The most essential pieces of a Kafka client are implemented, and we aim to provide all relevant consumer, producer, and admin APIs.

• Project Scope
• Installation
• Usage
  • Consuming Messages
  • Producing Messages
• Higher Level Libraries
  • Message Processing Frameworks
  • Message Publishing Libraries
• Forking
• Development
• Example
• Versions

While rdkafka-ruby aims to simplify the use of librdkafka in Ruby applications, it's important to understand the limitations of this library:

• No Complex Producers/Consumers: This library does not intend to offer complex producers or consumers. The aim is to stick closely to the functionalities provided by librdkafka itself.

• Focus on librdkafka Capabilities: Features that can be achieved directly in Ruby, without specific needs from librdkafka, are outside the scope of this library.

• Existing High-Level Functionalities: Certain high-level functionalities like producer metadata cache and simple consumer are already part of the library. Although they fall slightly outside the primary goal, they will remain part of the contract, given their existing usage.

When installed, this gem downloads and compiles librdkafka. If you have any problems installing the gem, please open an issue.

Please see the documentation for full details on how to use this gem. Below are two quick examples.

Unless you are seeking specific low-level capabilities, we strongly recommend using Karafka and WaterDrop when working with Kafka. These are higher-level libraries also maintained by us based on rdkafka-ruby.

Subscribe to a topic and get messages. Kafka will automatically spread the available partitions over consumers with the same group id.

config = {
  :"bootstrap.servers" => "localhost:9092",
  :"group.id" => "ruby-test"
}
consumer = Rdkafka::Config.new(config).consumer
consumer.subscribe("ruby-test-topic")

consumer.each do |message|
  puts "Message received: #{message}"
end

Produce several messages, put the delivery handles in an array, and wait for them before exiting. This way the messages will be batched and efficiently sent to Kafka.

config = {:"bootstrap.servers" => "localhost:9092"}
producer = Rdkafka::Config.new(config).producer
delivery_handles = []

100.times do |i|
  puts "Producing message #{i}"
  delivery_handles << producer.produce(
      topic:   "ruby-test-topic",
      payload: "Payload #{i}",
      key:     "Key #{i}"
  )
end

delivery_handles.each(&:wait)

Note that creating a producer consumes some resources that will not be released until it #close is explicitly called, so be sure to call Config#producer only as necessary.

Currently, there are two actively developed frameworks based on rdkafka-ruby, that provide higher-level API that can be used to work with Kafka messages and one library for publishing messages.

• Karafka - Ruby and Rails efficient Kafka processing framework.
• Racecar - A simple framework for Kafka consumers in Ruby

• WaterDrop – Standalone Karafka library for producing Kafka messages.

When working with rdkafka-ruby, it's essential to know that the underlying librdkafka library does not support fork-safe operations, even though it is thread-safe. Forking a process after initializing librdkafka clients can lead to unpredictable behavior due to inherited file descriptors and memory states. This limitation requires careful handling, especially in Ruby applications that rely on forking.

To address this, it's highly recommended to:

• Never initialize any rdkafka-ruby producers or consumers before forking to avoid state corruption.
• Before forking, always close any open producers or consumers if you've opened any.
• Use high-level libraries like WaterDrop and Karafka, which provide abstractions for handling librdkafka's intricacies.

Contributors are encouraged to focus on enhancements that align with the core goal of the library. We appreciate contributions but will likely not accept pull requests for features that:

• Implement functionalities that can achieved using standard Ruby capabilities without changes to the underlying rdkafka-ruby bindings.
• Deviate significantly from the primary aim of providing librdkafka bindings with Ruby-friendly interfaces.

A Docker Compose file is included to run Kafka. To run that:

docker-compose up

Run bundle and cd ext && bundle exec rake && cd .. to download and compile librdkafka.

You can then run bundle exec rspec to run the tests. To see rdkafka debug output:

DEBUG_PRODUCER=true bundle exec rspec
DEBUG_CONSUMER=true bundle exec rspec

After running the tests, you can bring the cluster down to start with a clean slate:

docker-compose down

To see everything working, run these in separate tabs:

bundle exec rake consume_messages
bundle exec rake produce_messages