New announcements for Serverless, Network, RUM, and more from Dash! New announcements from Dash!

Tracing Ruby Applications

ddtrace is Datadog’s tracing client for Ruby. It is used to trace requests as they flow across web servers, databases and microservices so that developers have high visibility into bottlenecks and troublesome requests.

Getting started

For the general APM documentation, see our setup documentation.

For more information about what APM looks like once your application is sending information to Datadog, take a look at Visualizing your APM data.

To contribute, check out the contribution guidelines and development guide.

Table of Contents

Compatibility

Supported Ruby interpreters:

TypeDocumentationVersionSupport typeGem version support
MRIhttps://www.ruby-lang.org/2.6FullLatest
2.5FullLatest
2.4FullLatest
2.3FullLatest
2.2FullLatest
2.1FullLatest
2.0FullLatest
1.9.3Maintenance (until August 6th, 2020)< 0.27.0
1.9.1Maintenance (until August 6th, 2020)< 0.27.0
JRubyhttp://jruby.org/9.1.5AlphaLatest

Supported web servers:

TypeDocumentationVersionSupport type
Pumahttp://puma.io/2.16+ / 3.6+Full
Unicornhttps://bogomips.org/unicorn/4.8+ / 5.1+Full
Passengerhttps://www.phusionpassenger.com/5.0+Full

Supported tracing frameworks:

TypeDocumentationVersionGem version support
OpenTracinghttps://github.com/opentracing/opentracing-ruby0.4.1+ (w/ Ruby 2.1+)>= 0.16.0

Full support indicates all tracer features are available.

Deprecated indicates support will transition to Maintenance in a future release.

Maintenance indicates only critical bugfixes are backported until EOL.

EOL indicates support is no longer provided.

Installation

The following steps will help you quickly start tracing your Ruby application.

Setup the Datadog Agent

Before downloading tracing on your application, install the Datadog Agent. The Ruby APM tracer sends trace data through the Datadog Agent.

Install and configure the Datadog Agent, see additional documentation for tracing Docker applications.

Quickstart for Rails applications

  1. Add the ddtrace gem to your Gemfile:

    source 'https://rubygems.org'
    gem 'ddtrace'
  2. Install the gem with bundle install

  3. Create a config/initializers/datadog.rb file containing:

    Datadog.configure do |c|
      # This will activate auto-instrumentation for Rails
      c.use :rails
    end

    You can also activate additional integrations here (see Integration instrumentation)

Quickstart for Ruby applications

  1. Install the gem with gem install ddtrace
  2. Add a configuration block to your Ruby application:

    require 'ddtrace'
    Datadog.configure do |c|
      # Configure the tracer here.
      # Activate integrations, change tracer settings, etc...
      # By default without additional configuration, nothing will be traced.
    end
  3. Add or activate instrumentation by doing either of the following:

Quickstart for OpenTracing

  1. Install the gem with gem install ddtrace
  2. To your OpenTracing configuration file, add the following:

    require 'opentracing'
    require 'ddtrace'
    require 'ddtrace/opentracer'
    
    # Activate the Datadog tracer for OpenTracing
    OpenTracing.global_tracer = Datadog::OpenTracer::Tracer.new
  3. (Optional) Add a configuration block to your Ruby application to configure Datadog with:

    Datadog.configure do |c|
      # Configure the Datadog tracer here.
      # Activate integrations, change tracer settings, etc...
      # By default without additional configuration,
      # no additional integrations will be traced, only
      # what you have instrumented with OpenTracing.
    end
  4. (Optional) Add or activate additional instrumentation by doing either of the following:

Final steps for installation

After setting up, your services will appear on the APM services page within a few minutes. Learn more about using the APM UI.

Manual Instrumentation

If you aren’t using a supported framework instrumentation, you may want to manually instrument your code.

To trace any Ruby code, you can use the Datadog.tracer.trace method:

Datadog.tracer.trace(name, options) do |span|
  # Wrap this block around the code you want to instrument
  # Additionally, you can modify the span here.
  # e.g. Change the resource name, set tags, etc...
end

Where name should be a String that describes the generic kind of operation being done (e.g. 'web.request', or 'request.parse')

And options is an optional Hash that accepts the following parameters:

KeyTypeDescriptionDefault
serviceStringThe service name which this span belongs (e.g. 'my-web-service')Tracer default-service, $PROGRAM_NAME or 'ruby'
resourceStringName of the resource or action being operated on. Traces with the same resource value will be grouped together for the purpose of metrics (but still independently viewable.) Usually domain specific, such as a URL, query, request, etc. (e.g. 'Article#submit', http://example.com/articles/list.)name of Span.
span_typeStringThe type of the span (such as 'http', 'db', etc.)nil
child_ofDatadog::Span / Datadog::ContextParent for this span. If not provided, will automatically become current active span.nil
start_timeIntegerWhen the span actually starts. Useful when tracing events that have already happened.Time.now.utc
tagsHashExtra tags which should be added to the span.{}
on_errorProcHandler invoked when a block is provided to trace, and it raises an error. Provided span and error as arguments. Sets error on the span by default.`proc {

It’s highly recommended you set both service and resource at a minimum. Spans without a service or resource as nil will be discarded by the Datadog agent.

Example of manual instrumentation in action:

get '/posts' do
  Datadog.tracer.trace('web.request', service: 'my-blog', resource: 'GET /posts') do |span|
    # Trace the activerecord call
    Datadog.tracer.trace('posts.fetch') do
      @posts = Posts.order(created_at: :desc).limit(10)
    end

    # Add some APM tags
    span.set_tag('http.method', request.request_method)
    span.set_tag('posts.count', @posts.length)

    # Trace the template rendering
    Datadog.tracer.trace('template.render') do
      erb :index
    end
  end
end

Asynchronous tracing

It might not always be possible to wrap Datadog.tracer.trace around a block of code. Some event or notification based instrumentation might only notify you when an event begins or ends.

To trace these operations, you can trace code asynchronously by calling Datadog.tracer.trace without a block:

# Some instrumentation framework calls this after an event finishes...
def db_query(start, finish, query)
  span = Datadog.tracer.trace('database.query')
  span.resource = query
  span.start_time = start
  span.finish(finish)
end

Calling Datadog.tracer.trace without a block will cause the function to return a Datadog::Span that is started, but not finished. You can then modify this span however you wish, then close it finish.

You must not leave any unfinished spans. If any spans are left open when the trace completes, the trace will be discarded. You can activate debug mode to check for warnings if you suspect this might be happening.

To avoid this scenario when handling start/finish events, you can use Datadog.tracer.active_span to get the current active span.

# e.g. ActiveSupport::Notifications calls this when an event starts
def start(name, id, payload)
  # Start a span
  Datadog.tracer.trace(name)
end

# e.g. ActiveSupport::Notifications calls this when an event finishes
def finish(name, id, payload)
  # Retrieve current active span (thread-safe)
  current_span = Datadog.tracer.active_span
  unless current_span.nil?
    current_span.resource = payload[:query]
    current_span.finish
  end
end

Enriching traces from nested methods

You can tag additional information to the current active span from any method. Note however that if the method is called and there is no span currently active active_span will be nil.

# e.g. adding tag to active span

current_span = Datadog.tracer.active_span
current_span.set_tag('my_tag', 'my_value') unless current_span.nil?

You can also get the root span of the current active trace using the active_root_span method. This method will return nil if there is no active trace.

# e.g. adding tag to active root span

current_root_span = Datadog.tracer.active_root_span
current_root_span.set_tag('my_tag', 'my_value') unless current_root_span.nil?

Integration instrumentation

Many popular libraries and frameworks are supported out-of-the-box, which can be auto-instrumented. Although they are not activated automatically, they can be easily activated and configured by using the Datadog.configure API:

Datadog.configure do |c|
  # Activates and configures an integration
  c.use :integration_name, options
end

options is a Hash of integration-specific configuration settings.

For a list of available integrations, and their configuration options, please refer to the following:

NameKeyVersions SupportedHow to configureGem source
Action Viewaction_view>= 3.2, < 6.0LinkLink
Active Model Serializersactive_model_serializers>= 0.9LinkLink
Action Packaction_pack>= 3.2, < 6.0LinkLink
Active Recordactive_record>= 3.2, < 6.0LinkLink
Active Supportactive_support>= 3.2, < 6.0LinkLink
AWSaws>= 2.0LinkLink
Concurrent Rubyconcurrent_ruby>= 0.9LinkLink
Dallidalli>= 2.7LinkLink
DelayedJobdelayed_job>= 4.1LinkLink
Elastic Searchelasticsearch>= 6.0LinkLink
Ethonethon>= 0.11.0LinkLink
Exconexcon>= 0.62LinkLink
Faradayfaraday>= 0.14LinkLink
Grapegrape>= 1.0LinkLink
GraphQLgraphql>= 1.7.9LinkLink
gRPCgrpc>= 1.10LinkLink
MongoDBmongo>= 2.0LinkLink
MySQL2mysql2>= 0.3.21LinkLink
Net/HTTPhttp(Any supported Ruby)LinkLink
Racecarracecar>= 0.3.5LinkLink
Rackrack>= 1.4.7LinkLink
Railsrails>= 3.2, < 6.0LinkLink
Rakerake>= 12.0LinkLink
Redisredis>= 3.2, < 4.0LinkLink
Resqueresque>= 1.0, < 2.0LinkLink
Rest Clientrest-client>= 1.8LinkLink
Sequelsequel>= 3.41LinkLink
Shoryukenshoryuken>= 4.0.2LinkLink
Sidekiqsidekiq>= 3.5.4LinkLink
Sinatrasinatra>= 1.4.5LinkLink
Sucker Punchsucker_punch>= 2.0LinkLink

Action View

Most of the time, Active Support is set up as part of Rails, but it can be activated separately:

require 'actionview'
require 'ddtrace'

Datadog.configure do |c|
  c.use :action_view, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for rendering instrumentation.action_view
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer
template_base_pathUsed when the template name is parsed. If you don’t store your templates in the views/ folder, you may need to change this value'views/'

Active Model Serializers

The Active Model Serializers integration traces the serialize event for version 0.9+ and the render event for version 0.10+.

require 'active_model_serializers'
require 'ddtrace'

Datadog.configure do |c|
  c.use :active_model_serializers, options
end

my_object = MyModel.new(name: 'my object')
ActiveModelSerializers::SerializableResource.new(test_obj).serializable_hash
KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for active_model_serializers instrumentation.'active_model_serializers'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Action Pack

Most of the time, Action Pack is set up as part of Rails, but it can be activated separately:

require 'actionpack'
require 'ddtrace'

Datadog.configure do |c|
  c.use :action_pack, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for rendering instrumentation.action_pack
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Active Record

Most of the time, Active Record is set up as part of a web framework (Rails, Sinatra…) however, it can be set up alone:

require 'tmpdir'
require 'sqlite3'
require 'active_record'
require 'ddtrace'

Datadog.configure do |c|
  c.use :active_record, options
end

Dir::Tmpname.create(['test', '.sqlite']) do |db|
  conn = ActiveRecord::Base.establish_connection(adapter: 'sqlite3',
                                                 database: db)
  conn.connection.execute('SELECT 42') # traced!
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to the global setting, false for off.false
orm_service_nameService name used for the Ruby ORM portion of active_record instrumentation. Overrides service name for ORM spans if explicitly set, which otherwise inherit their service from their parent.'active_record'
service_nameService name used for database portion of active_record instrumentation.Name of database adapter (e.g. 'mysql2')
tracerDatadog::Tracer used to perform instrumentation. Usually, you don’t need to set this.Datadog.tracer

Configuring trace settings per database

You can configure trace settings per database connection by using the describes option:

# Provide a `:describes` option with a connection key.
# Any of the following keys are acceptable and equivalent to one another.
# If a block is provided, it yields a Settings object that
# accepts any of the configuration options listed above.

Datadog.configure do |c|
  # Symbol matching your database connection in config/database.yml
  # Only available if you are using Rails with ActiveRecord.
  c.use :active_record, describes: :secondary_database, service_name: 'secondary-db'

  c.use :active_record, describes: :secondary_database do |second_db|
    second_db.service_name = 'secondary-db'
  end

  # Connection string with the following connection settings:
  # Adapter, user, host, port, database
  c.use :active_record, describes: 'mysql2://root@127.0.0.1:3306/mysql', service_name: 'secondary-db'

  # Hash with following connection settings
  # Adapter, user, host, port, database
  c.use :active_record, describes: {
      adapter:  'mysql2',
      host:     '127.0.0.1',
      port:     '3306',
      database: 'mysql',
      username: 'root'
    },
    service_name: 'secondary-db'
end

If ActiveRecord traces an event that uses a connection that matches a key defined by describes, it will use the trace settings assigned to that connection. If the connection does not match any of the described connections, it will use default settings defined by c.use :active_record instead.

Active Support

Most of the time, Active Support is set up as part of Rails, but it can be activated separately:

require 'activesupport'
require 'ddtrace'

Datadog.configure do |c|
  c.use :active_support, options
end

cache = ActiveSupport::Cache::MemoryStore.new
cache.read('city')

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
cache_serviceService name used for caching with active_support instrumentation.active_support-cache
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

AWS

The AWS integration will trace every interaction (e.g. API calls) with AWS services (S3, ElastiCache etc.).

require 'aws-sdk'
require 'ddtrace'

Datadog.configure do |c|
  c.use :aws, options
end

# Perform traced call
Aws::S3::Client.new.list_buckets

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for aws instrumentation'aws'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Concurrent Ruby

The Concurrent Ruby integration adds support for context propagation when using ::Concurrent::Future. Making sure that code traced within the Future#execute will have correct parent set.

To activate your integration, use the Datadog.configure method:

# Inside Rails initializer or equivalent
Datadog.configure do |c|
  # Patches ::Concurrent::Future to use ExecutorService that propagates context
  c.use :concurrent_ruby, options
end

# Pass context into code executed within Concurrent::Future
Datadog.tracer.trace('outer') do
  Concurrent::Future.execute { Datadog.tracer.trace('inner') { } }.wait
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
service_nameService name used for concurrent-ruby instrumentation'concurrent-ruby'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Dalli

Dalli integration will trace all calls to your memcached server:

require 'dalli'
require 'ddtrace'

# Configure default Dalli tracing behavior
Datadog.configure do |c|
  c.use :dalli, options
end

# Configure Dalli tracing behavior for single client
client = Dalli::Client.new('localhost:11211', options)
client.set('abc', 123)

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for dalli instrumentation'memcached'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

DelayedJob

The DelayedJob integration uses lifecycle hooks to trace the job executions.

You can enable it through Datadog.configure:

require 'ddtrace'

Datadog.configure do |c|
  c.use :delayed_job, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for DelayedJob instrumentation'delayed_job'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

The Elasticsearch integration will trace any call to perform_request in the Client object:

require 'elasticsearch/transport'
require 'ddtrace'

Datadog.configure do |c|
  c.use :elasticsearch, options
end

# Perform a query to ElasticSearch
client = Elasticsearch::Client.new url: 'http://127.0.0.1:9200'
response = client.perform_request 'GET', '_cluster/health'

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
quantizeHash containing options for quantization. May include :show with an Array of keys to not quantize (or :all to skip quantization), or :exclude with Array of keys to exclude entirely.{}
service_nameService name used for elasticsearch instrumentation'elasticsearch'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Ethon

The ethon integration will trace any HTTP request through Easy or Multi objects. Note that this integration also supports Typhoeus library which is based on Ethon.

require 'ddtrace'

Datadog.configure do |c|
  c.use :ethon, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
distributed_tracingEnables distributed tracingtrue
service_nameService name for ethon instrumentation.'ethon'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Excon

The excon integration is available through the ddtrace middleware:

require 'excon'
require 'ddtrace'

# Configure default Excon tracing behavior
Datadog.configure do |c|
  c.use :excon, options
end

connection = Excon.new('https://example.com')
connection.get

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
distributed_tracingEnables distributed tracingtrue
error_handlerA Proc that accepts a response parameter. If it evaluates to a truthy value, the trace span is marked as an error. By default only sets 5XX responses as errors.nil
service_nameService name for Excon instrumentation. When provided to middleware for a specific connection, it applies only to that connection object.'excon'
split_by_domainUses the request domain as the service name when set to true.false
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Configuring connections to use different settings

If you use multiple connections with Excon, you can give each of them different settings by configuring their constructors with middleware:

# Wrap the Datadog tracing middleware around the default middleware stack
Excon.new(
  'http://example.com',
  middlewares: Datadog::Contrib::Excon::Middleware.with(options).around_default_stack
)

# Insert the middleware into a custom middleware stack.
# NOTE: Trace middleware must be inserted after ResponseParser!
Excon.new(
  'http://example.com',
  middlewares: [
    Excon::Middleware::ResponseParser,
    Datadog::Contrib::Excon::Middleware.with(options),
    Excon::Middleware::Idempotent
  ]
)

Where options is a Hash that contains any of the parameters listed in the table above.

Faraday

The faraday integration is available through the ddtrace middleware:

require 'faraday'
require 'ddtrace'

# Configure default Faraday tracing behavior
Datadog.configure do |c|
  c.use :faraday, options
end

# Configure Faraday tracing behavior for single connection
connection = Faraday.new('https://example.com') do |builder|
  builder.use(:ddtrace, options)
  builder.adapter Faraday.default_adapter
end

connection.get('/foo')

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
distributed_tracingEnables distributed tracingtrue
error_handlerA Proc that accepts a response parameter. If it evaluates to a truthy value, the trace span is marked as an error. By default only sets 5XX responses as errors.nil
service_nameService name for Faraday instrumentation. When provided to middleware for a specific connection, it applies only to that connection object.'faraday'
split_by_domainUses the request domain as the service name when set to true.false
tracerDatadog::Tracer used to perform instrumentation. Usually, you don’t need to set this.Datadog.tracer

Grape

The Grape integration adds the instrumentation to Grape endpoints and filters. This integration can work side by side with other integrations like Rack and Rails.

To activate your integration, use the Datadog.configure method before defining your Grape application:

# api.rb
require 'grape'
require 'ddtrace'

Datadog.configure do |c|
  c.use :grape, options
end

# Then define your application
class RackTestingAPI < Grape::API
  desc 'main endpoint'
  get :success do
    'Hello world!'
  end
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.nil
enabledDefines whether Grape should be traced. Useful for temporarily disabling tracing. true or falsetrue
service_nameService name used for grape instrumentation'grape'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

GraphQL

The GraphQL integration activates instrumentation for GraphQL queries.

To activate your integration, use the Datadog.configure method:

# Inside Rails initializer or equivalent
Datadog.configure do |c|
  c.use :graphql, schemas: [YourSchema], options
end

# Then run a GraphQL query
YourSchema.execute(query, variables: {}, context: {}, operation_name: nil)

The use :graphql method accepts the following parameters. Additional options can be substituted in for options:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.nil
service_nameService name used for graphql instrumentation'ruby-graphql'
schemasRequired. Array of GraphQL::Schema objects which to trace. Tracing will be added to all the schemas listed, using the options provided to this configuration. If you do not provide any, then tracing will not be activated.[]
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Manually configuring GraphQL schemas

If you prefer to individually configure the tracer settings for a schema (e.g. you have multiple schemas with different service names), in the schema definition, you can add the following using the GraphQL API:

YourSchema = GraphQL::Schema.define do
  use(
    GraphQL::Tracing::DataDogTracing,
    service: 'graphql'
  )
end

Or you can modify an already defined schema:

YourSchema.define do
  use(
    GraphQL::Tracing::DataDogTracing,
    service: 'graphql'
  )
end

Do NOT use :graphql in Datadog.configure if you choose to configure manually, as to avoid double tracing. These two means of configuring GraphQL tracing are considered mutually exclusive.

gRPC

The grpc integration adds both client and server interceptors, which run as middleware before executing the service’s remote procedure call. As gRPC applications are often distributed, the integration shares trace information between client and server.

To setup your integration, use the Datadog.configure method like so:

require 'grpc'
require 'ddtrace'

Datadog.configure do |c|
  c.use :grpc, options
end

# Server side
server = GRPC::RpcServer.new
server.add_http2_port('localhost:50051', :this_port_is_insecure)
server.handle(Demo)
server.run_till_terminated

# Client side
client = Demo.rpc_stub_class.new('localhost:50051', :this_channel_is_insecure)
client.my_endpoint(DemoMessage.new(contents: 'hello!'))

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for grpc instrumentation'grpc'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Configuring clients to use different settings

In situations where you have multiple clients calling multiple distinct services, you may pass the Datadog interceptor directly, like so

configured_interceptor = Datadog::Contrib::GRPC::DatadogInterceptor::Client.new do |c|
  c.service_name = "Alternate"
end

alternate_client = Demo::Echo::Service.rpc_stub_class.new(
  'localhost:50052',
  :this_channel_is_insecure,
  :interceptors => [configured_interceptor]
)

The integration will ensure that the configured_interceptor establishes a unique tracing setup for that client instance.

MongoDB

The integration traces any Command that is sent from the MongoDB Ruby Driver to a MongoDB cluster. By extension, Object Document Mappers (ODM) such as Mongoid are automatically instrumented if they use the official Ruby driver. To activate the integration, simply:

require 'mongo'
require 'ddtrace'

Datadog.configure do |c|
  c.use :mongo, options
end

# Create a MongoDB client and use it as usual
client = Mongo::Client.new([ '127.0.0.1:27017' ], :database => 'artists')
collection = client[:people]
collection.insert_one({ name: 'Steve' })

# In case you want to override the global configuration for a certain client instance
Datadog.configure(client, options)

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
quantizeHash containing options for quantization. May include :show with an Array of keys to not quantize (or :all to skip quantization), or :exclude with Array of keys to exclude entirely.{ show: [:collection, :database, :operation] }
service_nameService name used for mongo instrumentation'mongodb'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

MySQL2

The MySQL2 integration traces any SQL command sent through mysql2 gem.

require 'mysql2'
require 'ddtrace'

Datadog.configure do |c|
  c.use :mysql2, options
end

client = Mysql2::Client.new(:host => "localhost", :username => "root")
client.query("SELECT * FROM users WHERE group='x'")

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for mysql2 instrumentation'mysql2'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Net/HTTP

The Net/HTTP integration will trace any HTTP call using the standard lib Net::HTTP module.

require 'net/http'
require 'ddtrace'

Datadog.configure do |c|
  c.use :http, options
end

Net::HTTP.start('127.0.0.1', 8080) do |http|
  request = Net::HTTP::Get.new '/index'
  response = http.request(request)
end

content = Net::HTTP.get(URI('http://127.0.0.1/index.html'))

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
distributed_tracingEnables distributed tracingtrue
service_nameService name used for http instrumentation'net/http'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

If you wish to configure each connection object individually, you may use the Datadog.configure as it follows:

client = Net::HTTP.new(host, port)
Datadog.configure(client, options)

Racecar

The Racecar integration provides tracing for Racecar jobs.

You can enable it through Datadog.configure:

require 'ddtrace'

Datadog.configure do |c|
  c.use :racecar, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for racecar instrumentation'racecar'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Rack

The Rack integration provides a middleware that traces all requests before they reach the underlying framework or application. It responds to the Rack minimal interface, providing reasonable values that can be retrieved at the Rack level.

This integration is automatically activated with web frameworks like Rails. If you’re using a plain Rack application, enable the integration it to your config.ru:

# config.ru example
require 'ddtrace'

Datadog.configure do |c|
  c.use :rack, options
end

use Datadog::Contrib::Rack::TraceMiddleware

app = proc do |env|
  [ 200, {'Content-Type' => 'text/plain'}, ['OK'] ]
end

run app

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.nil
applicationYour Rack application. Required for middleware_names.nil
distributed_tracingEnables distributed tracing so that this service trace is connected with a trace of another service if tracing headers are receivedtrue
headersHash of HTTP request or response headers to add as tags to the rack.request. Accepts request and response keys with Array values e.g. ['Last-Modified']. Adds http.request.headers.* and http.response.headers.* tags respectively.{ response: ['Content-Type', 'X-Request-ID'] }
middleware_namesEnable this if you want to use the middleware classes as the resource names for rack spans. Requires application option to use.false
quantizeHash containing options for quantization. May include :query or :fragment.{}
quantize.queryHash containing options for query portion of URL quantization. May include :show or :exclude. See options below. Option must be nested inside the quantize option.{}
quantize.query.showDefines which values should always be shown. Shows no values by default. May be an Array of strings, or :all to show all values. Option must be nested inside the query option.nil
quantize.query.excludeDefines which values should be removed entirely. Excludes nothing by default. May be an Array of strings, or :all to remove the query string entirely. Option must be nested inside the query option.nil
quantize.fragmentDefines behavior for URL fragments. Removes fragments by default. May be :show to show URL fragments. Option must be nested inside the quantize option.nil
request_queuingTrack HTTP request time spent in the queue of the frontend server. See HTTP request queuing for setup details. Set to true to enable.false
service_nameService name used for rack instrumentation'rack'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer
web_service_nameService name for frontend server request queuing spans. (e.g. 'nginx')'web-server'

Configuring URL quantization behavior

Datadog.configure do |c|
  # Default behavior: all values are quantized, fragment is removed.
  # http://example.com/path?category_id=1&sort_by=asc#featured --> http://example.com/path?category_id&sort_by
  # http://example.com/path?categories[]=1&categories[]=2 --> http://example.com/path?categories[]

  # Show values for any query string parameter matching 'category_id' exactly
  # http://example.com/path?category_id=1&sort_by=asc#featured --> http://example.com/path?category_id=1&sort_by
  c.use :rack, quantize: { query: { show: ['category_id'] } }

  # Show all values for all query string parameters
  # http://example.com/path?category_id=1&sort_by=asc#featured --> http://example.com/path?category_id=1&sort_by=asc
  c.use :rack, quantize: { query: { show: :all } }

  # Totally exclude any query string parameter matching 'sort_by' exactly
  # http://example.com/path?category_id=1&sort_by=asc#featured --> http://example.com/path?category_id
  c.use :rack, quantize: { query: { exclude: ['sort_by'] } }

  # Remove the query string entirely
  # http://example.com/path?category_id=1&sort_by=asc#featured --> http://example.com/path
  c.use :rack, quantize: { query: { exclude: :all } }

  # Show URL fragments
  # http://example.com/path?category_id=1&sort_by=asc#featured --> http://example.com/path?category_id&sort_by#featured
  c.use :rack, quantize: { fragment: :show }
end

Rails

The Rails integration will trace requests, database calls, templates rendering, and cache read/write/delete operations. The integration makes use of the Active Support Instrumentation, listening to the Notification API so that any operation instrumented by the API is traced.

To enable the Rails instrumentation, create an initializer file in your config/initializers folder:

# config/initializers/datadog.rb
require 'ddtrace'

Datadog.configure do |c|
  c.use :rails, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to the global setting, false for off.nil
cache_serviceCache service name used when tracing cache activity'<app_name>-cache'
controller_serviceService name used when tracing a Rails action controller'<app_name>'
database_serviceDatabase service name used when tracing database activity'<app_name>-<adapter_name>'
distributed_tracingEnables distributed tracing so that this service trace is connected with a trace of another service if tracing headers are receivedtrue
exception_controllerClass or Module which identifies a custom exception controller class. Tracer provides improved error behavior when it can identify custom exception controllers. By default, without this option, it ‘guesses’ what a custom exception controller looks like. Providing this option aids this identification.nil
middlewareAdd the trace middleware to the Rails application. Set to false if you don’t want the middleware to load.true
middleware_namesEnables any short-circuited middleware requests to display the middleware name as a resource for the trace.false
service_nameService name used when tracing application requests (on the rack level)'<app_name>' (inferred from your Rails application namespace)
template_base_pathUsed when the template name is parsed. If you don’t store your templates in the views/ folder, you may need to change this value'views/'
tracerDatadog::Tracer used to perform instrumentation. Usually, you don’t need to set this.Datadog.tracer

Supported versions

Ruby VersionsSupported Rails Versions
2.03.0 - 3.2
2.13.0 - 4.2
2.2 - 2.33.0 - 5.2
2.4 - 2.54.2.8 - 5.2
2.65.0 - 5.2

Rake

You can add instrumentation around your Rake tasks by activating the rake integration. Each task and its subsequent subtasks will be traced.

To activate Rake task tracing, add the following to your Rakefile:

# At the top of your Rakefile:
require 'rake'
require 'ddtrace'

Datadog.configure do |c|
  c.use :rake, options
end

task :my_task do
  # Do something task work here...
end

Rake::Task['my_task'].invoke

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to the global setting, false for off.false
enabledDefines whether Rake tasks should be traced. Useful for temporarily disabling tracing. true or falsetrue
quantizeHash containing options for quantization of task arguments. See below for more details and examples.{}
service_nameService name used for rake instrumentation'rake'
tracerDatadog::Tracer used to perform instrumentation. Usually, you don’t need to set this.Datadog.tracer

Configuring task quantization behavior

Datadog.configure do |c|
  # Given a task that accepts :one, :two, :three...
  # Invoked with 'foo', 'bar', 'baz'.

  # Default behavior: all arguments are quantized.
  # `rake.invoke.args` tag  --> ['?']
  # `rake.execute.args` tag --> { one: '?', two: '?', three: '?' }
  c.use :rake

  # Show values for any argument matching :two exactly
  # `rake.invoke.args` tag  --> ['?']
  # `rake.execute.args` tag --> { one: '?', two: 'bar', three: '?' }
  c.use :rake, quantize: { args: { show: [:two] } }

  # Show all values for all arguments.
  # `rake.invoke.args` tag  --> ['foo', 'bar', 'baz']
  # `rake.execute.args` tag --> { one: 'foo', two: 'bar', three: 'baz' }
  c.use :rake, quantize: { args: { show: :all } }

  # Totally exclude any argument matching :three exactly
  # `rake.invoke.args` tag  --> ['?']
  # `rake.execute.args` tag --> { one: '?', two: '?' }
  c.use :rake, quantize: { args: { exclude: [:three] } }

  # Remove the arguments entirely
  # `rake.invoke.args` tag  --> ['?']
  # `rake.execute.args` tag --> {}
  c.use :rake, quantize: { args: { exclude: :all } }
end

Redis

The Redis integration will trace simple calls as well as pipelines.

require 'redis'
require 'ddtrace'

Datadog.configure do |c|
  c.use :redis, options
end

# Perform Redis commands
redis = Redis.new
redis.set 'foo', 'bar'

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for redis instrumentation'redis'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

You can also set per-instance configuration as it follows:

customer_cache = Redis.new
invoice_cache = Redis.new

Datadog.configure(customer_cache, service_name: 'customer-cache')
Datadog.configure(invoice_cache, service_name: 'invoice-cache')

# Traced call will belong to `customer-cache` service
customer_cache.get(...)
# Traced call will belong to `invoice-cache` service
invoice_cache.get(...)

Resque

The Resque integration uses Resque hooks that wraps the perform method.

To add tracing to a Resque job:

require 'ddtrace'

class MyJob
  def self.perform(*args)
    # do_something
  end
end

Datadog.configure do |c|
  c.use :resque, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to the global setting, false for off.false
service_nameService name used for resque instrumentation'resque'
tracerDatadog::Tracer used to perform instrumentation. Usually, you don’t need to set this.Datadog.tracer
workersAn array including all worker classes you want to trace (e.g. [MyJob])[]

Rest Client

The rest-client integration is available through the ddtrace middleware:

require 'rest_client'
require 'ddtrace'

Datadog.configure do |c|
  c.use :rest_client, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
distributed_tracingEnables distributed tracingtrue
service_nameService name for rest_client instrumentation.'rest_client'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Sequel

The Sequel integration traces queries made to your database.

require 'sequel'
require 'ddtrace'

# Connect to database
database = Sequel.sqlite

# Create a table
database.create_table :articles do
  primary_key :id
  String :name
end

Datadog.configure do |c|
  c.use :sequel, options
end

# Perform a query
articles = database[:articles]
articles.all

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name for sequel instrumentationName of database adapter (e.g. 'mysql2')
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Only Ruby 2.0+ is supported.

Configuring databases to use different settings

If you use multiple databases with Sequel, you can give each of them different settings by configuring their respective Sequel::Database objects:

sqlite_database = Sequel.sqlite
postgres_database = Sequel.connect('postgres://user:password@host:port/database_name')

# Configure each database with different service names
Datadog.configure(sqlite_database, service_name: 'my-sqlite-db')
Datadog.configure(postgres_database, service_name: 'my-postgres-db')

Shoryuken

The Shoryuken integration is a server-side middleware which will trace job executions.

You can enable it through Datadog.configure:

require 'ddtrace'

Datadog.configure do |c|
  c.use :shoryuken, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for shoryuken instrumentation'shoryuken'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Sidekiq

The Sidekiq integration is a client-side & server-side middleware which will trace job queuing and executions respectively.

You can enable it through Datadog.configure:

require 'ddtrace'

Datadog.configure do |c|
  c.use :sidekiq, options
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
client_service_nameService name used for client-side sidekiq instrumentation'sidekiq-client'
service_nameService name used for server-side sidekiq instrumentation'sidekiq'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Sinatra

The Sinatra integration traces requests and template rendering.

To start using the tracing client, make sure you import ddtrace and use :sinatra after either sinatra or sinatra/base, and before you define your application/routes:

require 'sinatra'
require 'ddtrace'

Datadog.configure do |c|
  c.use :sinatra, options
end

get '/' do
  'Hello world!'
end

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.nil
distributed_tracingEnables distributed tracing so that this service trace is connected with a trace of another service if tracing headers are receivedtrue
headersHash of HTTP request or response headers to add as tags to the sinatra.request. Accepts request and response keys with Array values e.g. ['Last-Modified']. Adds http.request.headers.* and http.response.headers.* tags respectively.{ response: ['Content-Type', 'X-Request-ID'] }
resource_script_namesPrepend resource names with script namefalse
service_nameService name used for sinatra instrumentation'sinatra'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Sucker Punch

The sucker_punch integration traces all scheduled jobs:

require 'ddtrace'

Datadog.configure do |c|
  c.use :sucker_punch, options
end

# Execution of this job is traced
LogJob.perform_async('login')

Where options is an optional Hash that accepts the following parameters:

KeyDescriptionDefault
analytics_enabledEnable analytics for spans produced by this integration. true for on, nil to defer to global setting, false for off.false
service_nameService name used for sucker_punch instrumentation'sucker_punch'
tracerDatadog::Tracer used to perform instrumentation. Usually you don’t need to set this.Datadog.tracer

Advanced configuration

Tracer settings

To change the default behavior of the Datadog tracer, you can provide custom options inside the Datadog.configure block as in:

# config/initializers/datadog-tracer.rb

Datadog.configure do |c|
  c.tracer option_name: option_value, ...
end

Available options are:

  • enabled: defines if the tracer is enabled or not. If set to false the code could be still instrumented because of other settings, but no spans are sent to the local trace agent.
  • debug: set to true to enable debug logging.
  • hostname: set the hostname of the trace agent.
  • port: set the port the trace agent is listening on.
  • env: set the environment. Rails users may set it to Rails.env to use their application settings.
  • tags: set global tags that should be applied to all spans. Defaults to an empty hash
  • log: defines a custom logger.
  • partial_flush: set to true to enable partial trace flushing (for long running traces.) Disabled by default. Experimental.

Custom logging

By default, all logs are processed by the default Ruby logger. When using Rails, you should see the messages in your application log file.

Datadog client log messages are marked with [ddtrace] so you should be able to isolate them from other messages.

Additionally, it is possible to override the default logger and replace it by a custom one. This is done using the log attribute of the tracer.

f = File.new("my-custom.log", "w+")           # Log messages should go there
Datadog.configure do |c|
  c.tracer log: Logger.new(f)                 # Overriding the default tracer
end

Datadog::Tracer.log.info { "this is typically called by tracing code" }

Environment and tags

By default, the trace agent (not this library, but the program running in the background collecting data from various clients) uses the tags set in the agent config file, see our environments tutorial for details.

These values can be overridden at the tracer level:

Datadog.configure do |c|
  c.tracer tags: { 'env' => 'prod' }
end

This enables you to set this value on a per tracer basis, so you can have for example several applications reporting for different environments on the same host.

Ultimately, tags can be set per span, but env should typically be the same for all spans belonging to a given trace.

Sampling

ddtrace can perform trace sampling. While the trace agent already samples traces to reduce bandwidth usage, client sampling reduces the performance overhead.

Datadog::RateSampler samples a ratio of the traces. For example:

# Sample rate is between 0 (nothing sampled) to 1 (everything sampled).
sampler = Datadog::RateSampler.new(0.5) # sample 50% of the traces
Datadog.configure do |c|
  c.tracer sampler: sampler
end

Priority sampling

Priority sampling decides whether to keep a trace by using a priority attribute propagated for distributed traces. Its value indicates to the Agent and the backend about how important the trace is.

The sampler can set the priority to the following values:

  • Datadog::Ext::Priority::AUTO_REJECT: the sampler automatically decided to reject the trace.
  • Datadog::Ext::Priority::AUTO_KEEP: the sampler automatically decided to keep the trace.

Priority sampling is enabled by default. Enabling it ensures that your sampled distributed traces will be complete. Once enabled, the sampler will automatically assign a priority of 0 or 1 to traces, depending on their service and volume.

You can also set this priority manually to either drop a non-interesting trace or to keep an important one. For that, set the context#sampling_priority to:

  • Datadog::Ext::Priority::USER_REJECT: the user asked to reject the trace.
  • Datadog::Ext::Priority::USER_KEEP: the user asked to keep the trace.

When not using distributed tracing, you may change the priority at any time, as long as the trace incomplete. But it has to be done before any context propagation (fork, RPC calls) to be useful in a distributed context. Changing the priority after the context has been propagated causes different parts of a distributed trace to use different priorities. Some parts might be kept, some parts might be rejected, and this can cause the trace to be partially stored and remain incomplete.

If you change the priority, we recommend you do it as soon as possible - when the root span has just been created.

# First, grab the active span
span = Datadog.tracer.active_span

# Indicate to reject the trace
span.context.sampling_priority = Datadog::Ext::Priority::USER_REJECT

# Indicate to keep the trace
span.context.sampling_priority = Datadog::Ext::Priority::USER_KEEP

Distributed Tracing

Distributed tracing allows traces to be propagated across multiple instrumented applications so that a request can be presented as a single trace, rather than a separate trace per service.

To trace requests across application boundaries, the following must be propagated between each application:

PropertyTypeDescription
Trace IDIntegerID of the trace. This value should be the same across all requests that belong to the same trace.
Parent Span IDIntegerID of the span in the service originating the request. This value will always be different for each request within a trace.
Sampling PriorityIntegerSampling priority level for the trace. This value should be the same across all requests that belong to the same trace.

Such propagation can be visualized as:

Service A:
  Trace ID:  100000000000000001
  Parent ID: 0
  Span ID:   100000000000000123
  Priority:  1

  |
  | Service B Request:
  |   Metadata:
  |     Trace ID:  100000000000000001
  |     Parent ID: 100000000000000123
  |     Priority:  1
  |
  V

Service B:
  Trace ID:  100000000000000001
  Parent ID: 100000000000000123
  Span ID:   100000000000000456
  Priority:  1

  |
  | Service C Request:
  |   Metadata:
  |     Trace ID:  100000000000000001
  |     Parent ID: 100000000000000456
  |     Priority:  1
  |
  V

Service C:
  Trace ID:  100000000000000001
  Parent ID: 100000000000000456
  Span ID:   100000000000000789
  Priority:  1

Via HTTP

For HTTP requests between instrumented applications, this trace metadata is propagated by use of HTTP Request headers:

PropertyTypeHTTP Header name
Trace IDIntegerx-datadog-trace-id
Parent Span IDIntegerx-datadog-parent-id
Sampling PriorityIntegerx-datadog-sampling-priority

Such that:

Service A:
  Trace ID:  100000000000000001
  Parent ID: 0
  Span ID:   100000000000000123
  Priority:  1

  |
  | Service B HTTP Request:
  |   Headers:
  |     x-datadog-trace-id:          100000000000000001
  |     x-datadog-parent-id:         100000000000000123
  |     x-datadog-sampling-priority: 1
  |
  V

Service B:
  Trace ID:  100000000000000001
  Parent ID: 100000000000000123
  Span ID:   100000000000000456
  Priority:  1

  |
  | Service C HTTP Request:
  |   Headers:
  |     x-datadog-trace-id:          100000000000000001
  |     x-datadog-parent-id:         100000000000000456
  |     x-datadog-sampling-priority: 1
  |
  V

Service C:
  Trace ID:  100000000000000001
  Parent ID: 100000000000000456
  Span ID:   100000000000000789
  Priority:  1

Activating distributed tracing for integrations

Many integrations included in ddtrace support distributed tracing. Distributed tracing is enabled by default, but can be activated via configuration settings.

  • If your application receives requests from services with distributed tracing activated, you must activate distributed tracing on the integrations that handle these requests (e.g. Rails)
  • If your application send requests to services with distributed tracing activated, you must activate distributed tracing on the integrations that send these requests (e.g. Faraday)
  • If your application both sends and receives requests implementing distributed tracing, it must activate all integrations that handle these requests.

For more details on how to activate distributed tracing for integrations, see their documentation:

Using the HTTP propagator

To make the process of propagating this metadata easier, you can use the Datadog::HTTPPropagator module.

On the client:

Datadog.tracer.trace('web.call') do |span|
  # Inject span context into headers (`env` must be a Hash)
  Datadog::HTTPPropagator.inject!(span.context, env)
end

On the server:

Datadog.tracer.trace('web.work') do |span|
  # Build a context from headers (`env` must be a Hash)
  context = HTTPPropagator.extract(request.env)
  Datadog.tracer.provider.context = context if context.trace_id
end

HTTP request queuing

Traces that originate from HTTP requests can be configured to include the time spent in a frontend web server or load balancer queue before the request reaches the Ruby application.

This functionality is experimental and deactivated by default.

To activate this feature, you must add an X-Request-Start or X-Queue-Start header from your web server (i.e., Nginx). The following is an Nginx configuration example:

# /etc/nginx/conf.d/ruby_service.conf
server {
    listen 8080;

    location / {
      proxy_set_header X-Request-Start "t=${msec}";
      proxy_pass http://web:3000;
    }
}

Then you must enable the request queuing feature in the integration handling the request.

For Rack-based applications, see the documentation for details for enabling this feature.

Processing Pipeline

Some applications might require that traces be altered or filtered out before they are sent upstream. The processing pipeline allows users to create processors to define such behavior.

Processors can be any object that responds to #call accepting trace as an argument (which is an Array of Datadog::Spans.)

For example:

lambda_processor = ->(trace) do
  # Processing logic...
  trace
end

class MyCustomProcessor
  def call(trace)
    # Processing logic...
    trace
  end
end
custom_processor = MyFancyProcessor.new

#call blocks of processors must return the trace object; this return value will be passed to the next processor in the pipeline.

These processors must then be added to the pipeline via Datadog::Pipeline.before_flush:

Datadog::Pipeline.before_flush(lambda_processor, custom_processor)

You can also define processors using the short-hand block syntax for Datadog::Pipeline.before_flush:

Datadog::Pipeline.before_flush do |trace|
  trace.delete_if { |span| span.name =~ /forbidden/ }
end

Filtering

You can use the Datadog::Pipeline::SpanFilter processor to remove spans, when the block evaluates as truthy:

Datadog::Pipeline.before_flush(
  # Remove spans that match a particular resource
  Datadog::Pipeline::SpanFilter.new { |span| span.resource =~ /PingController/ },
  # Remove spans that are trafficked to localhost
  Datadog::Pipeline::SpanFilter.new { |span| span.get_tag('host') == 'localhost' }
)

Processing

You can use the Datadog::Pipeline::SpanProcessor processor to modify spans:

Datadog::Pipeline.before_flush(
  # Strip matching text from the resource field
  Datadog::Pipeline::SpanProcessor.new { |span| span.resource.gsub!(/password=.*/, '') }
)

Trace correlation

In many cases, such as logging, it may be useful to correlate trace IDs to other events or data streams, for easier cross-referencing. The tracer can produce a correlation identifier for the currently active trace via active_correlation, which can be used to decorate these other data sources.

# When a trace is active...
Datadog.tracer.trace('correlation.example') do
  # Returns #<Datadog::Correlation::Identifier>
  correlation = Datadog.tracer.active_correlation
  correlation.trace_id # => 5963550561812073440
  correlation.span_id # => 2232727802607726424
end

# When a trace isn't active...
correlation = Datadog.tracer.active_correlation
# Returns #<Datadog::Correlation::Identifier>
correlation = Datadog.tracer.active_correlation
correlation.trace_id # => 0
correlation.span_id # => 0

After setting up Lograge in a Rails application, modify the custom_options block in your environment configuration file (e.g. config/environments/production.rb) to add the trace IDs:

config.lograge.custom_options = lambda do |event|
  # Retrieves trace information for current thread
  correlation = Datadog.tracer.active_correlation

  {
    # Adds IDs as tags to log output
    :dd => {
      :trace_id => correlation.trace_id,
      :span_id => correlation.span_id
    },
    :ddsource => ["ruby"],
    :params => event.payload[:params].reject { |k| %w(controller action).include? k }
  }
end

For logging in Rails applications

Rails applications which are configured with an ActiveSupport::TaggedLogging logger can append correlation IDs as tags to log output. The default Rails logger implements this tagged logging, making it easier to add correlation tags.

In your Rails environment configuration file, add the following:

Rails.application.configure do
  config.log_tags = [proc { Datadog.tracer.active_correlation.to_s }]
end

# Web requests will produce:
# [dd.trace_id=7110975754844687674 dd.span_id=7518426836986654206] Started GET "/articles" for 172.22.0.1 at 2019-01-16 18:50:57 +0000
# [dd.trace_id=7110975754844687674 dd.span_id=7518426836986654206] Processing by ArticlesController#index as */*
# [dd.trace_id=7110975754844687674 dd.span_id=7518426836986654206]   Article Load (0.5ms)  SELECT "articles".* FROM "articles"
# [dd.trace_id=7110975754844687674 dd.span_id=7518426836986654206] Completed 200 OK in 7ms (Views: 5.5ms | ActiveRecord: 0.5ms)

For logging in Ruby applications

To add correlation IDs to your logger, add a log formatter which retrieves the correlation IDs with Datadog.tracer.active_correlation, then add them to the message.

To properly correlate with Datadog logging, be sure the following is present in the log message:

  • dd.trace_id=<TRACE_ID>: Where <TRACE_ID> is equal to Datadog.tracer.active_correlation.trace_id or 0 if no trace is active during logging.
  • dd.span_id=<SPAN_ID>: Where <SPAN_ID> is equal to Datadog.tracer.active_correlation.span_id or 0 if no trace is active during logging.

By default, Datadog::Correlation::Identifier#to_s will return dd.trace_id=<TRACE_ID> dd.span_id=<SPAN_ID>.

An example of this in practice:

require 'ddtrace'
require 'logger'

logger = Logger.new(STDOUT)
logger.progname = 'my_app'
logger.formatter  = proc do |severity, datetime, progname, msg|
  "[#{datetime}][#{progname}][#{severity}][#{Datadog.tracer.active_correlation}] #{msg}\n"
end

# When no trace is active
logger.warn('This is an untraced operation.')
# [2019-01-16 18:38:41 +0000][my_app][WARN][dd.trace_id=0 dd.span_id=0] This is an untraced operation.

# When a trace is active
Datadog.tracer.trace('my.operation') { logger.warn('This is a traced operation.') }
# [2019-01-16 18:38:41 +0000][my_app][WARN][dd.trace_id=8545847825299552251 dd.span_id=3711755234730770098] This is a traced operation.

Configuring the transport layer

By default, the tracer submits trace data using Net::HTTP to 127.0.0.1:8126, the default location for the Datadog trace agent process. However, the tracer can be configured to send its trace data to alternative destinations, or by alternative protocols.

Some basic settings, such as hostname and port, can be configured using tracer settings.

Using the Net::HTTP adapter

The Net adapter submits traces using Net::HTTP over TCP. It is the default transport adapter.

Datadog.configure do |c|
  c.tracer transport_options: proc do |t|
    # Hostname, port, and additional options. :timeout is in seconds.
    t.adapter :net_http, '127.0.0.1', 8126, { timeout: 1 }
  }
end

Using the Unix socket adapter

The UnixSocket adapter submits traces using Net::HTTP over Unix socket.

To use, first configure your trace agent to listen by Unix socket, then configure the tracer with:

Datadog.configure do |c|
  c.tracer transport_options: proc { |t|
    # Provide filepath to trace agent Unix socket
    t.adapter :unix, '/tmp/ddagent/trace.sock'
  }
end

Using the transport test adapter

The Test adapter is a no-op transport that can optionally buffer requests. For use in test suites or other non-production environments.

Datadog.configure do |c|
  c.tracer transport_options: proc { |t|
    # Set transport to no-op mode. Does not retain traces.
    t.adapter :test

    # Alternatively, you can provide a buffer to examine trace output.
    # The buffer must respond to '<<'.
    t.adapter :test, []
  }
end

Using a custom transport adapter

Custom adapters can be configured with:

Datadog.configure do |c|
  c.tracer transport_options: proc { |t|
    # Initialize and pass an instance of the adapter
    custom_adapter = CustomAdapter.new
    t.adapter custom_adapter
  }
end

Metrics

The tracer and its integrations can produce some additional metrics that can provide useful insight into the performance of your application. These metrics are collected with dogstatsd-ruby, and can be sent to the same Datadog agent to which you send your traces.

To configure your application for metrics collection:

  1. Configure your Datadog agent for StatsD
  2. Add gem 'dogstatsd-ruby' to your Gemfile

For application runtime

If runtime metrics are configured, the trace library will automatically collect and send metrics about the health of your application.

To configure runtime metrics, add the following configuration:

# config/initializers/datadog.rb
require 'datadog/statsd'
require 'ddtrace'

Datadog.configure do |c|
  # To enable runtime metrics collection, set `true`. Defaults to `false`
  # You can also set DD_RUNTIME_METRICS_ENABLED=true to configure this.
  c.runtime_metrics_enabled = true

  # Optionally, you can configure the Statsd instance used for sending runtime metrics.
  # Statsd is automatically configured with default settings if `dogstatsd-ruby` is available.
  # You can configure with host and port of Datadog agent; defaults to 'localhost:8125'.
  c.runtime_metrics statsd: Datadog::Statsd.new
end

See the Dogstatsd documentation for more details about configuring Datadog::Statsd.

The stats sent will include:

NameTypeDescription
runtime.ruby.class_countgaugeNumber of classes in memory space.
runtime.ruby.thread_countgaugeNumber of threads.
runtime.ruby.gc.*.gaugeGarbage collection statistics (one per value in GC.stat)

In addition, all metrics include the following tags:

NameDescription
languageProgramming language traced. (e.g. ruby)
serviceList of services this associated with this metric.

OpenTracing

For setting up Datadog with OpenTracing, see out Quickstart for OpenTracing section for details.

Configuring Datadog tracer settings

The underlying Datadog tracer can be configured by passing options (which match Datadog::Tracer) when configuring the global tracer:

# Where `options` is a Hash of options provided to Datadog::Tracer
OpenTracing.global_tracer = Datadog::OpenTracer::Tracer.new(options)

It can also be configured by using Datadog.configure described in the Tracer settings section.

Activating and configuring integrations

By default, configuring OpenTracing with Datadog will not automatically activate any additional instrumentation provided by Datadog. You will only receive spans and traces from OpenTracing instrumentation you have in your application.

However, additional instrumentation provided by Datadog can be activated alongside OpenTracing using Datadog.configure, which can be used to enhance your tracing further. To activate this, see Integration instrumentation for more details.

Supported serialization formats

TypeSupported?Additional information
OpenTracing::FORMAT_TEXT_MAPYes
OpenTracing::FORMAT_RACKYesBecause of the loss of resolution in the Rack format, please note that baggage items with names containing either upper case characters or - will be converted to lower case and _ in a round-trip respectively. We recommend avoiding these characters or accommodating accordingly on the receiving end.
OpenTracing::FORMAT_BINARYNo