aws-xray

The unofficial AWS X-Ray Tracing SDK for Ruby. It enables you to capture in-coming HTTP requests and out-going HTTP requests and send them to X-Ray daemon automatically.

AWS X-Ray is a distributed tracing system. See more detail about AWS X-Ray at official document. If you want to know what is distributed tracing, what is problems behind, etc.., please refer Google's Dapper paper.

aws-xray has full features to build and send tracing data to AWS X-Ray.

• Propagation support in both single and multi thread environments.
• Instrumentation for major libraries.
• Recording HTTP requests/responses and errors.
• Annotation and metadata support.
• Sampling.

• net/http
• rack
• faraday
• rsolr

Just require aws/xray/rails. It uses your application name as a service name by default. e.g. Legacy::MyBlog -> legacy-my-blog.

# Gemfile
gem 'aws-xray', require: ['aws/xray/rails', 'aws/xray/hooks/net_http']

Requiring aws/xray/rails inserts Rack middleware to the middleware stack and the middleware automatically starts tracing context. Another requiring aws/xray/hooks/net_http inserts a hook to net/http and it records out-going HTTP requests/responses automatically.

Then setup X-Ray daemon in your runtime environment. Once the daemon is ready, run your application with some environment variables required by aws-xray gem.

• AWS_XRAY_LOCATION: Point to X-Ray daemon's bind address and port. e.g. localhost:2000.
• AWS_XRAY_SAMPLING_RATE: Set sampling rate. If you are just checking behavior, you can disable sampling by setting 1.
• AWS_XRAY_EXCLUDED_PATHS: Set your application's health check paths to avoid tracing health check requests.

You then see your application builds and sends tracing data to X-Ray daemon.

Recommend setting these operatinal concern via environment variables.

See more configuration at API documentation.

aws-xray does not send any trace data by default. Set AWS_XRAY_LOCATION environment variable like AWS_XRAY_LOCATION=localhost:2000.

In container environments, we run X-Ray daemon container beside application container. For that case, pass AWS_XRAY_LOCATION environment variable to container to specify host and port of X-Ray daemon.

docker run --link xray:xray --env AWS_XRAY_LOCATION=xray:2000 my-application

Sampling rate should be a float within 0 to 1. Both 0 and 1 are acceptable. e.g. 0 means never sampled, 1 means always sampled, 0.3 means 30% of requests (or traces in non-Rack app) will be sampled. The default sampling rate is undefined so you should set your own sampling rate on production systems.

Set sampling rate with AWS_XRAY_SAMPLING_RATE env var.

To avoid tracing health checking requests, use "excluded paths" configuration.

• Environment variable: AWS_XRAY_EXCLUDED_PATHS=/health_check,/another_check
• Global configuration: Aws::Xray.config.excluded_paths = ['/health_check', '/another_check', %r{/token/.+}]

aws-xray automatically tries to set application version by reading app_root/REVISION file. If you want to set another version, set it with:

# In initialization phase.
Aws::Xray.config.version = 'deadbeef'

aws-xray records hostname by default.

If you want to record specific annotation in all of your segments, configure like:

Aws::Xray.config.default_annotation = Aws::Xray.config.default_annotation.merge(key: 'value')

Keys must be alphanumeric characters with underscore and values must be one of String or Integer or Boolean values.

For metadata:

Aws::Xray.config.default_metadata = Aws::Xray.config.default_metadata.merge(key: ['some', 'meaningful', 'value'])

Note: See official document about annotation and metadata in AWS X-Ray.

When aws-xray fails to send segments due to system call errors, it logs errors to stderr by default. If you want to track these errors, for example with Sentry, you can configure your own error handler:

Aws::Xray.config.segment_sending_error_handler = MyCustomErrorHandler.new

The error handler must be callable object and receive 2 arguments and 2 keyword arguments. See Aws::Xray::DefaultErrorHandler for more details.

Optionally, aws-xray offers an error handler which integrates with Sentry. To use it:

Aws::Xray.config.segment_sending_error_handler = Aws::Xray::ErrorHandlerWithSentry.new

Set Aws::Xray.config.record_caller_of_http_requests = true if you want to investigate the caller of specific HTTP requests. It records caller of net/http and Faraday middleware.

Tracing context is thread local. To pass current tracing context, copy current tracing context:

Thread.new(Aws::Xray.current_context.copy) do |context|
  Aws::Xray.with_given_context(context) do
    # Do something
  end
end

require 'aws/xray'
require 'aws/xray/hooks/net_http'

# Start new tracing context then perform arbitrary actions in the block.
Aws::Xray.trace(name: 'my-app-batch') do |seg|
  # Record out-going HTTP request/response with net/http hook.
  Net::HTTP.get('example.com', '/index.html')

  # Record arbitrary actions as subsegment.
  Aws::Xray.start_subsegment(name: 'fetch-user', remote: true) do |sub|
    # DB access or something to trace.
  end
end

# config.ru
require 'aws-xray'
Aws::Xray.config.name = 'my-app'
use Aws::Xray::Rack

This enables your app to start tracing context.

require 'aws/xray/faraday'
Faraday.new('...', headers: { 'Host' => 'down-stream-app-id' } ) do |builder|
  builder.use Aws::Xray::Faraday
  # ...
end

If you don't use any Service Discovery tools, pass the down stream app name to the Faraday middleware:

require 'aws/xray/faraday'
Faraday.new('...') do |builder|
  builder.use Aws::Xray::Faraday, 'down-stream-app-id'
  # ...
end

You can enable all the hooks with:

# Gemfile
gem 'aws-xray', require: 'aws/xray/hooks/all'

To monkey patch net/http and records out-going http requests automatically, just require aws/xray/hooks/net_http:

If you can pass headers for net/http client, you can setup subsegment name via X-Aws-Xray-Name header:

Net::HTTP.start(host, port) do |http|
  req = Net::HTTP::Get.new(uri, { 'X-Aws-Xray-Name' => 'target-app' })
  http.request(req)
end

If you can't access headers, e.g. external client library like aws-sdk or dogapi-rb, setup subsegment name by Aws::Xray.overwrite:

client = Aws::Sns::Client.new
response = Aws::Xray.overwrite(name: 'sns') do
  client.create_topic(...)
end

When you want to name solr requests, use this hook by require aws/xray/hooks/rsolr. The typical usecase is you use local haproxy to proxy to solr instances and you want to distinguish these requests from other reqeusts using local haproxy.

If you want to give a specific name, configure it:

Aws::Xray.config.solr_hook_name = 'solr-development'