ruby-amqp/hutch

Hutch is a Ruby library for enabling asynchronous inter-service communication in a service-oriented architecture, using RabbitMQ.

To install with RubyGems:

gem install hutch

• Requirements
• Overview
  • Project Maturity
• Consumers
  • Message Processing Tracers
• Running Hutch
  • Loading Consumers
  • Stopping Hutch
• Producers
  • Producer Configuration
  • Publisher Confirms
  • Writing Well-Behaved Publishers
• Configuration
  • Config File
  • Environment variables
  • Configuration precedence
  • Generated list of configuration options

• Hutch requires Ruby 2.4+ or JRuby 9K.
• Hutch requires RabbitMQ 3.3 or later.

Hutch is a conventions-based framework for writing services that communicate over RabbitMQ. Hutch is opinionated: it uses topic exchanges for message distribution and makes some assumptions about how consumers and publishers should work.

With Hutch, consumers are stored in separate files and include the Hutch::Consumer module. They are then loaded by a command line runner which connects to RabbitMQ, sets up queues and bindings, and so on. Publishers connect to RabbitMQ via Hutch.connect and publish using Hutch.publish.

Hutch uses Bunny or March Hare (on JRuby) under the hood.

Hutch is a mature project that was originally extracted from production systems at GoCardless in 2013 and is now maintained by its contributors and users.

Consumers receive messages from a RabbitMQ queue. That queue may be bound to one or more topics (represented by routing keys).

To create a consumer, include the Hutch::Consumer module in a class that defines a #process method. #process should take a single argument, which will be a Message object. The Message object encapsulates the message data, along with any associated metadata. To access properties of the message, use Hash-style indexing syntax:

message[:id]  # => "02ABCXYZ"

To subscribe to a topic, pass a routing key to consume in the class definition. To bind to multiple routing keys, simply pass extra routing keys in as additional arguments. Refer to the RabbitMQ docs on topic exchanges for more information about how to use routing keys. Here's an example consumer:

class FailedPaymentConsumer
  include Hutch::Consumer
  consume 'gc.ps.payment.failed'

  def process(message)
    mark_payment_as_failed(message[:id])
  end
end

By default, the queue name will be named using the consumer class. You can set the queue name explicitly by using the queue_name method:

class FailedPaymentConsumer
  include Hutch::Consumer
  consume 'gc.ps.payment.failed'
  queue_name 'failed_payments'

  def process(message)
    mark_payment_as_failed(message[:id])
  end
end

It is possible to set some custom options to consumer's queue explicitly. This example sets the consumer's queue as a quorum queue and to operate in the lazy mode. The initial_group_size argument is optional.

class FailedPaymentConsumer
  include Hutch::Consumer
  consume 'gc.ps.payment.failed'
  queue_name 'failed_payments'
  lazy_queue
  quorum_queue initial_group_size: 3

  def process(message)
    mark_payment_as_failed(message[:id])
  end
end

You can also set custom arguments per consumer. This example declares a consumer with a maximum length of 10 messages:

class FailedPaymentConsumer
  include Hutch::Consumer
  consume 'gc.ps.payment.failed'
  arguments 'x-max-length' => 10
end

This sets the x-max-length header. For more details, see the RabbitMQ documentation on Queue Length Limit. To find out more about custom queue arguments, consult the RabbitMQ documentation on AMQP Protocol Extensions.

Consumers can write to Hutch's log by calling the logger method. The logger method returns a Logger object.

class FailedPaymentConsumer
  include Hutch::Consumer
  consume 'gc.ps.payment.failed'

  def process(message)
    logger.info "Marking payment #{message[:id]} as failed"
    mark_payment_as_failed(message[:id])
  end
end

If you are using Hutch with Rails and want to make Hutch log to the Rails logger rather than stdout, add this to config/initializers/hutch.rb

Hutch::Logging.logger = Rails.logger

A logger can be set for the client by adding this config before calling Hutch.connect

client_logger = Logger.new("/path/to/bunny.log")
Hutch::Config.set(:client_logger, client_logger)

See this RabbitMQ tutorial on topic exchanges to learn more.

Tracers allow you to track message processing.

This will enable NewRelic custom instrumentation:

Hutch::Config.set(:tracer, Hutch::Tracers::NewRelic)

And this will enable Datadog custom instrumentation:

Hutch::Config.set(:tracer, Hutch::Tracers::Datadog)

Batteries included!

After installing the Hutch gem, you should be able to start it by simply running hutch on the command line. hutch takes a number of options:

$ hutch -h
usage: hutch [options]
        --mq-host HOST               Set the RabbitMQ host
        --mq-port PORT               Set the RabbitMQ port
    -t, --[no-]mq-tls                Use TLS for the AMQP connection
        --mq-tls-cert FILE           Certificate  for TLS client verification
        --mq-tls-key FILE            Private key for TLS client verification
        --mq-exchange EXCHANGE       Set the RabbitMQ exchange
        --mq-vhost VHOST             Set the RabbitMQ vhost
        --mq-username USERNAME       Set the RabbitMQ username
        --mq-password PASSWORD       Set the RabbitMQ password
        --mq-api-host HOST           Set the RabbitMQ API host
        --mq-api-port PORT           Set the RabbitMQ API port
    -s, --[no-]mq-api-ssl            Use SSL for the RabbitMQ API
        --config FILE                Load Hutch configuration from a file
        --require PATH               Require a Rails app or path
        --[no-]autoload-rails        Require the current rails app directory
    -q, --quiet                      Quiet logging
    -v, --verbose                    Verbose logging
        --version                    Print the version and exit
    -h, --help                       Show this message and exit

The first three are for configuring which RabbitMQ instance to connect to. --require is covered in the next section. Configurations can also be specified in a YAML file for convenience by passing the file location to the --config option. The file should look like:

mq_username: peter
mq_password: rabbit
mq_host: broker.yourhost.com

Passing a setting as a command-line option will overwrite what's specified in the config file, allowing for easy customization.

Using Hutch with a Rails app is simple. Either start Hutch in the working directory of a Rails app, or pass the path to a Rails app in with the --require option. Consumers defined in Rails apps should be placed with in the app/consumers/ directory, to allow them to be auto-loaded when Rails boots.

If you're using the new Zeitwerk autoloader (enabled by default in Rails 6) and the consumers are not loaded in development environment you will need to trigger the autoloading in an initializer with

::Zeitwerk::Loader.eager_load_all

or with something more specific like

autoloader = Rails.autoloaders.main

Dir.glob(File.join('app/consumers', '*_consumer.rb')).each do |consumer|
  autoloader.preload(consumer)
end

It is possible to load only a subset of consumers. This is done by defining a consumer group under the consumer_groups configuration key:

consumer_groups:
  payments:
    - DepositConsumer
    - CashoutConsumer
  notification:
    - EmailNotificationConsumer

To only load a group of consumers, use the --only-group option:

hutch --only-group=payments --config=/path/to/hutch.yaml

To require files that define consumers manually, simply pass each file as an option to --require. Hutch will automatically detect whether you've provided a Rails app or a standard file, and take the appropriate behaviour:

# loads a rails app
hutch --require path/to/rails-app
# loads a ruby file
hutch --require path/to/file.rb

Hutch supports graceful stops. That means that if done correctly, Hutch will wait for your consumer to finish processing before exiting.

To gracefully stop your workers, you may send the following signals to your Hutch processes: INT, TERM, or QUIT.

kill -SIGINT 123 # or kill -2 123
kill -SIGTERM 456 # or kill -15 456
kill -SIGQUIT 789 # or kill -3 789

Hutch includes a publish method for sending messages to Hutch consumers. When possible, this should be used, rather than directly interfacing with RabbitMQ libraries.

Hutch.connect
Hutch.publish('routing.key', subject: 'payment', action: 'received')

Producers are not run with the 'hutch' command. You can specify configuration options as follows:

Hutch::Config.set(:mq_exchange, 'name')

For maximum message reliability when producing messages, you can force Hutch to use Publisher Confirms and wait for a confirmation after every message published. This is the safest possible option for publishers but also results in a significant throughput drop.

Hutch::Config.set(:force_publisher_confirms, true)

You may need to send messages to Hutch from languages other than Ruby. This prevents the use of Hutch.publish, requiring custom publication code to be written. There are a few things to keep in mind when writing producers that send messages to Hutch.

• Make sure that the producer exchange name matches the exchange name that Hutch is using.
• Hutch works with topic exchanges, check the producer is also using topic exchanges.
• Use message routing keys that match those used in your Hutch consumers.
• Be sure your exchanges are marked as durable. In the Ruby AMQP gem, this is done by passing durable: true to the exchange creation method.
• Publish messages as persistent.
• Using publisher confirms is highly recommended.

Here's an example of a well-behaved publisher, minus publisher confirms:

AMQP.connect(host: config[:host]) do |connection|
  channel  = AMQP::Channel.new(connection)
  exchange = channel.topic(config[:exchange], durable: true)

  message = JSON.dump({ subject: 'Test', id: 'abc' })
  exchange.publish(message, routing_key: 'test', persistent: true)
end

If using publisher confirms with amqp gem, see this issue and this gist for more info.

It is recommended to use a separate config file, unless you use URIs for connection (see below).

Known configuration parameters are:

• mq_host: RabbitMQ hostname (default: localhost)
• mq_port: RabbitMQ port (default: 5672)
• mq_vhost: vhost to use (default: /)
• mq_username: username to use (default: guest, only can connect from localhost as of RabbitMQ 3.3.0)
• mq_password: password to use (default: guest)
• mq_tls: should TLS be used? (default: false)
• mq_tls_cert: path to client TLS certificate (public key)
• mq_tls_key: path to client TLS private key
• mq_tls_ca_certificates: array of paths to CA keys (if not specified to Hutch, will default to Bunny defaults which are system-dependent)
• mq_verify_peer: should SSL certificate be verified? (default: true)
• require_paths: array of paths to require
• autoload_rails: should Hutch command line runner try to automatically load Rails environment files?
• daemonise: should Hutch runner process daemonise?
• pidfile: path to PID file the runner should use
• channel_prefetch: basic.qos prefetch value to use (default: 0, no limit). See Bunny and RabbitMQ documentation.
• publisher_confirms: enables publisher confirms. Leaves it up to the app how they are tracked (e.g. using Hutch::Broker#confirm_select callback or Hutch::Broker#wait_for_confirms)
• force_publisher_confirms: enables publisher confirms, forces Hutch::Broker#wait_for_confirms for every publish. This is the safest option which also offers the lowest throughput.
• log_level: log level used by the standard Ruby logger (default: Logger::INFO)
• error_handlers: a list of error handler objects, see classes in Hutch::ErrorHandlers. All configured handlers will be invoked unconditionally in the order listed.
• error_acknowledgements: a chain of responsibility of objects that acknowledge/reject/requeue messages when an exception happens, see classes in Hutch::Acknowledgements.
• mq_exchange: exchange to use for publishing (default: hutch)
• heartbeat: RabbitMQ heartbeat timeout (default: 30)
• connection_timeout: Bunny's socket open timeout (default: 11)
• read_timeout: Bunny's socket read timeout (default: 11)
• write_timeout: Bunny's socket write timeout (default: 11)
• automatically_recover: Bunny's enable/disable network recovery (default: true)
• network_recovery_interval: Bunny's reconnect interval (default: 1)
• tracer: tracer to use to track message processing
• namespace: A namespace string to help group your queues (default: nil)

The file configuration options mentioned above can also be passed in via environment variables, using the HUTCH_ prefix, eg.

• connection_timeout → HUTCH_CONNECTION_TIMEOUT.

In order from lowest to highest precedence:

0. Default values
1. HUTCH_* environment variables
2. Configuration file
3. Explicit settings through Hutch::Config.set

Generate with

0. yard doc lib/hutch/config.rb
1. Copy the Configuration section from doc/Hutch/Config.html here, with the anchor tags stripped.