The SignalFx Tracing Library for Ruby helps you instrument Ruby applications with the OpenTracing API to capture and report distributed traces to SignalFx.
The library consists of an auto-instrumentor that works with OpenTracing community-provided instrumentations, and provides a bootstrap utility to help install instrumentations. It also configures and uses a Jaeger tracer to send trace data to SignalFx.
Here are the requirements and supported software for the library.
- MRI Ruby (CRuby) 2.0+
- Puma >= 3.0.0
- Passenger >= 5.0.25
Library | Instrumentation name | Versions Supported |
---|---|---|
ActiveRecord | signalfx-activerecord-opentracing | ~> 5.0 |
Elasticsearch | signalfx-elasticsearch-instrumentation | >= 6.0.2 |
Faraday | signalfx-faraday-instrumentation | >= 0.2.1 |
Grape | signalfx-grape-instrumentation | >= 0.2.0 |
Mongo | signalfgx-mongodb-instrumentation | >= 2.1.0 |
Mysql2 | signalfx-mysql2-instrumentation | >= 0.4.0 |
Net::HTTP | signalfx-nethttp-instrumentation | Ruby >= 2.0 |
Pg | signalfx-pg-instrumentation | >= 0.18.0 |
Rack | signalfx-rack-tracer | >= 0.10.0 |
Rails | signalfx-rails-instrumentation | >= 3.0.0 |
Redis | signalfx-redis-instrumentation | >= 4.0.0 |
RestClient | signalfx-restclient-instrumentation | >= 1.5.0 |
Sequel | signalfx-sequel-instrumentation | >= 3.47.0 |
Sidekiq | signalfx-sidekiq-opentracing | >= 0.7.0 |
Sinatra | signalfx-sinatra-instrumentation | >= 1.0.0 |
Instrumentation for routes using Puma or Passenger is provided through
Rack. If you use a framework that builds on top of Rack, such as Rails or
Sinatra, install the signalfx-rack-tracer
instrumentation with your dependency manager
or with the bootstrap utility. In these cases, the routes through the web
server are automatically traced.
Follow these steps to install the tracing library. You can either use the bootstrap utility to install the tracing library and its dependencies or manually install everything.
The bootstrap utility updates your Gemfile and installs the required dependencies for you. For information about the bootstrap utility, see the sfx-rb-trace-bootstrap file.
The steps assume you have RubyGems and Bundler.
- Install the tracing library:
$ gem install signalfx-tracing
- View the list of instrumentations you can install with the bootstrap utility:
$ sfx-rb-trace-bootstrap --list
- Use the bootstrap utility to install applicable instrumentations for your
application. For example, this is how you add Rails and Redis:
For information about instrumentation names, see supported libraries and their current versions in
$ sfx-rb-trace-bootstrap --install-deps rails,redis
gem.deps.rb
. If you configure Rails instrumentation, it also configures Active Record instrumentation, so you don't need to instrument both.
- Download the latest release of the tracing library.
- Add
signalfx-tracing
to your application's Gemfile:$ gem 'signalfx-tracing'
- Add each applicable instrumentation to your application's Gemfile. For
example, this is how you add Rails and Redis:
For information about instrumentation names, see supported libraries and their current versions in
$ gem 'signalfx-rails-instrumentation' $ gem 'signalfx-redis-instrumentation'
gem.deps.rb
. If you configure Rails instrumentation, it also configures Active Record instrumentation, so you don't need to instrument both. - Install the gems for the tracing library and instrumentations:
$ bundle install
Configure the instrumentation anywhere in the setup portion of your code or before importing and using any libraries that need to be traced.
For example, with Rails, configure instrumentation in config/initializer/tracing.rb
.
You can configure instrumentation automatically or manually. Manual instrumentation is convenient when you want to trace only some libraries.
If the default configuration values don't apply for your environment, override them before running the process you instrument.
configure parameter |
Environment variable | Default | Notes |
---|---|---|---|
tracer | N/A | nil |
The OpenTracing global tracer. |
ingest_url | SIGNALFX_ENDPOINT_URL | http://localhost:9080/v1/trace |
The endpoint the tracer sends spans to. Send spans to a Smart Agent, OpenTelemetry Collector, or a SignalFx ingest endpoint. |
service_name | SIGNALFX_SERVICE_NAME | signalfx-ruby-tracing |
The name to identify the service in SignalFx. |
access_token | SIGNALFX_ACCESS_TOKEN | '' |
The SignalFx organization access token. |
span_tags | SIGNALFX_SPAN_TAGS | nil |
Comma-separated list of tags included in every reported span. For example, "key1:val1,key2:val2". Use only string values for tags. |
N/A | SIGNALFX_RECORDED_VALUE_MAX_LENGTH | 1200 |
Maximum length an attribute value can have. Values longer than this are truncated. |
Configure the auto-instrumentor to check for modules defined in the code and instrument them if available:
require 'signalfx/tracing'
SignalFx::Tracing::Instrumenter.configure(auto_instrument:true)
Specify which libraries to instrument:
require 'signalfx/tracing'
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(<:myLibName>)
...
end
Here's information about instrumenting each supported library.
This instrumentation creates spans for each Active Record query using the Active Support notifications framework. If you configure Rails instrumentation, it also configures Active Record instrumentation, so you don't need to instrument both.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i activerecord
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:ActiveRecord)
end
Elasticsearch queries through the Ruby client are traced using a wrapper around the transport.
The forked source for the instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i elasticsearch
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Elasticsearch, auto_instrument: true)
end
When the auto_instrument
option is true
, it will allow all new Elasticsearch
clients to automatically wrap the default or configured transport and get
traces for queries. No additional configuration is necessary.
Alternatively, the instrumentation can be used selectively by setting a custom transport on the clients to be traced manually:
require 'elasticsearch'
require 'elasticsearch-tracer'
client = Elasticsearch::TracingClient.new
client.transport = Elasticsearch::Tracer::Transport.new(tracer: OpenTracing.global_tracer,
active_span: -> { OpenTracing.global_tracer.active_span },
transport: client.transport)
Faraday HTTP client instrumentation automatically creates spans for outgoing requests. If the remote service has instrumentation that is aware of Rack, those spans will be automatically nested with Faraday's spans.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i faraday
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Faraday)
end
To use the instrumentation directly without patching, the Faraday middleware must be inserted for each new connection:
conn = Faraday.new(url: 'http://localhost:3000/') do |faraday|
faraday.use Faraday::Tracer
end
For more detailed usage, please check the instrumentation's page.
This instrumentation subscribes to ActiveSupport notifications emitted by the
Grape API. It patches Grape::API
to automatically insert the Rack::Tracer
middleware and trace requests.
The source for this instrumentation is located here
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i grape
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Grape)
end
instrument
takes these optional arguments:
tracer
: custom tracer to use. Defaults toOpenTracing.global_tracer
parent_span
: parent span to group spans or block that returns a span. Default:nil
disable_patching
: disable patching if managing the middleware stack manually. Default:false
If patching is disabled, but spans nested by request are still desired, then the Rack middleware must be manually added to the API class.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i rack
require 'rack/tracer'
class MyAPI << Grape::API
insert 0, Rack::Tracer
...
end
Please see the instrumentation's page for more details.
Mongo driver instrumentation traces queries performed through the Ruby Mongodb driver.
The source for this instrumentation is located here
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i mongodb
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:MongoDB)
end
Mysql2 instrumentation traces all queries performed with the Mysql2 client.
The source for this instrumentation is located here
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i mysql2
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Mysql2)
end
This automatically traces all requests using Net::HTTP.
The source for this instrumentation is located here.
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:NetHttp, tracer: tracer)
end
An optional tracer
named argument can be provided to use a custom tracer.
It will default to OpenTracing.global_tracer
if not provided.
Pg instrumentation traces all queries performed with the pg client.
The source for this instrumentation is located here
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i pg
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:pg)
end
Rack spans are created using the signalfx-rack-tracer
gem. This is enabled
automatically for other frameworks that are built on top of Rack, but it can
also be separately enabled.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i rack
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Rack)
end
use Rack::Tracer
Rails applications can be traced using the notifications provided by ActiveSupport.
It will use signalfx-rack-tracer
to trace by requests.
The forked source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i rails
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Rails)
end
Optionally, to disable Rack instrumentation, set the rack_tracer
field to false.
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Rails, rack_tracer: false, exclude_events: [])
end
By default, all Rails ActiveSupport notifications are traced. However, if this
is too noisy, events to ignore can be passed as an array as exclude_events
.
A full list of events can be seen on the instrumentation's Readme.
Note that if rack_tracer
is set to false
, requests propagated to the Rails
app will not be extracted. For example, if a traced service makes a request to
an endpoint served by the Rails app, it will not be automatically nested.
This instrumentation traces commands performed using the Redis client.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i redis
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Redis, tracer: tracer)
end
Arguments:
tracer
: Optional custom tracer to use for this instrumentation- Default:
OpenTracing.global_tracer
- Default:
RestClient requests can be patched to automatically be wrapped in a span. It will also inject the span context so remote services can extract it.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i restclient
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:RestClient, tracer: tracer, propagate: true)
end
Arguments:
tracer
: Optional custom tracer to use for this instrumentation- Default:
OpenTracing.global_tracer
- Default:
propagate
: Propagate spans to the request endpoint.- Default:
false
- Default:
Sequel instrumentation adds extensions to the Database and Dataset to trace queries.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i sequel
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Sequel)
end
Arguments:
tracer
: Optional custom tracer for this instrumentation- Default:
OpenTracing.global_tracer
- Default:
Sidekiq instrumentation traces worker job submissions and execution via Sidekiq middleware. The instrumenter registers both client and server middleware that use job metadata to represent all job submissions and their invocations. Trace context propagation adds to this job metadata to unifiy distributed client and server requests and processing.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i sidekiq
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Sidekiq, propagate: false)
end
Arguments:
tracer
: Optional custom tracer for this instrumentation- Default:
OpenTracing.global_tracer
- Default:
propagate
: Optional boolean to enable/disable trace context injection via job metadata- Default:
true
- Default:
Sinatra instrumentation traces requests and template rendering. The instrumenter
registers a Sinatra extension that uses signalfx-rack-tracer
to trace requests and
monkey-patches to trace view rendering. Rack instrumentation is automatically
enabled when using Sinatra instrumentation.
The source for this instrumentation is located here.
$ # install the instrumentation if not done previously
$ sfx-rb-trace-bootstrap -i sinatra
SignalFx::Tracing::Instrumenter.configure do |p|
p.instrument(:Sinatra)
end
The logger, based on the Ruby Logger, can be configured by setting the following environment variables:
Environmental Variable Name | Description | Default |
---|---|---|
SIGNALFX_LOG_PATH |
The path to the desired file where the instrumentation logs will be written. Besides customized paths, output options also include STDOUT and STDERR . |
/var/log/signalfx/signalfx-ruby-tracing.log |
SIGNALFX_LOG_SHIFT_AGE |
The desired number of old log files to keep, or frequency of rotation. Options include: daily , weekly or monthly ) |
0 |
SIGNALFX_LOG_SHIFT_SIZE |
The desired maximum size of log files (this only applies when. A new one would be created when the maximum is reached. | 1048576 (1 MB) |
SIGNALFX_LOG_LEVEL |
The severity criteria for recording logs (from least to most severe). Options: debug , info , warn , error , fatal , unknown |
warn |
More information regarding the logging configuration may be found here.
NB:
- If the default path for
SIGNALFX_LOG_PATH
(that is,/var/log/signalfx/signalfx-ruby-tracing.log
) is to be used, then please create the directory and or file (if necessary) and grant the relevant access permissions to the instrumentation user process. If there are permission issues, the instrumentation will default to logging to the standard error (STDERR) handle, until the path is provided to which logs can be written without any access issues.