The RPM Contrib Gem
The rpm_contrib
gem contains instrumentation for the New Relic RPM agent
contributed by the community of RPM users. It requires the RPM Agent to run.
To use the rpm_contrib gem, install the rpm_contrib
gem from rubygems.org.
It will also install the required version of the newrelic_rpm
gem if it's not
already installed.
For Rails 3.0 and when using Bundler, add these dependencies to your Gemfile:
gem 'rpm_contrib'
gem 'newrelic_rpm'
For Rails 2.1 and later, add these dependencies to your in your environment.rb:
config.gem 'rpm_contrib'
config.gem 'newrelic_rpm'
For other frameworks, make sure you load rubygems if it isn't already, then just require the rpm_contrib gem:
require 'rubygems'
require 'rpm_contrib'
When you load the rpm_contrib gem, the newrelic_rpm
gem will also be
initialized. No need for a separate require statement for newrelic_rpm
.
In non-Rails frameworks, it's important that the New Relic Agent gets loaded as late as possible, or that the final initialization hook is called after all other frameworks have loaded:
DependencyDetection.detect!
Troubleshooting Startup
If you've set up your gems to load as described above and you are still not seeing
data in RPM, there may be a bug in detecting your framework. Try setting the
environment variable NEWRELIC_DISPATCHER
to the name of your app server (Camping,
Resque, Rake, etc), and please report to us if this fixes the problem so we can
fix the auto-detection logic.
If this does not help then set the log_level
to debug
in the newrelic.yml
file
and examine the newrelic_agent.log
file for errors after restarting your app.
Supported Frameworks
A number of frameworks are supported in the contrib gem. They are all turned on by default but you can add settings to your newrelic.yml to disable any of them.
ActiveMessaging
The gem will detect the underlying ActiveMessaging::Processor class and instrument the on_message
method
It can be disabled with the disable_active_messaging
flag in your newrelic.yml file.
Cassandra
The gem will instrument Cassandra so it should be visible in transaction traces and the web transactions page.
You can disable it with disable_cassandra_instrumentation
in your newrelic.yml file.
Camping
The gem will detect a Camping app but you need to manually add the instrumentation to your configuration file. See RPMContrib::Instrumentation::Camping for more information.
In addition you will need to load the gems in the following order: 1) Camping, 2) rpm_contrib, 3) newrelic_rpm.
Crack
The gem will instrument the Crack parsers for JSON and XML - you should see them in transaction traces and the web transactions page.
You can disable it with disable_crack
in your newrelic.yml file.
Curb
The gem will instrument both Curl::Easy and Curl::Multi - they should show up similarly to Net::HTTP in the UI
You can disable it with disable_curb
in your newrelic.yml file.
Elastic Search
The gem will instrument ElasticSearch::Client. The metrics should show up in the UI
You can disable it with disable_elastic_search_instrumentation
in your newrelic.yml file.
KyotoTycoon
The gem will instrument KyotoTycoon.
You can disable it with disable_kyototycoon
in your newrelic.yml file.
Paperclip
No special configuration required for Paperclip visibility.
You can disable it by setting disable_paperclip
to true in your newrelic.yml file.
Picky
The gem will instrument the Picky semantic search engine so it should be visible in transaction traces and the web transactions page.
You can disable it with disable_picky
in your newrelic.yml file.
MongoDB
Our instrumentation works on the underlying 'Mongo' library.
You can disable it by setting 'disable_mongodb' to true in your newrelic.yml file.
Resque
To instrument jobs you no longer need to have your Job class inherit from Resque::Job or include the Resque::Plugins::NewRelicInstrumentation module. The module definition was left in for backward compatibility.
To disable resque, set 'disable_resque' to true in your newrelic.yml file.
Redis
Redis instrumentation will record operations as well as allWeb
and allOther
summary metrics under the Database/Redis
metric namespace. This instrumentation
supports Redis versions 1.x and 2.x.
To disable Redis instrumentation, set 'disable_redis' to true in your newrelic.yml file.
Riak
RiakClient is instrumented. Its opereations are recorded under Database in the the response time graph.
To disable Riak instrumentation, set 'disable_riak_client' to true in your newrelic.yml file.
Ripple
Ripple is instrumented. Its opereations are recorded under Database in the the response time graph.
To disable Riak instrumentation, set 'disable_ripple' to true in your newrelic.yml file.
Sinatra view instrumentation
This adds instrumentation to the render
methods in Sinatra::Base
You can disable it with disable_sinatra_template
in your newrelic.yml file.
ThinkingSphinx instrumentation
This adds instrumentation to the initialize
and results
method of ThinkingSphinx::Search
You can disable it with disable_thinking_sphinx
in your newrelic.yml file.
Typhoeus instrumentation
This adds instrumentation to the Typhoeus::Request class for 'GET' requests
You can disable it with disable_typhoeus
in your newrelic.yml file.
Ultrasphinx instrumentation
This adds basic instrumentation to the run
and results
method of Ultrasphinx::Search
You can disable it with disable_ultrasphinx
in your newrelic.yml file.
Workling
This adds instrumentation to the Workling::Base and all children, for all defined public methods not inherited from the Workling::Base class
You can disable it with disable_workling
in your newrelic.yml file.
YAJL
This adds instrumentation to the YAJL json parser
You can disable it with disable_yajl_instrumentation
in your newrelic.yml file.
AWS/S3
Get metrics on how S3 is performing for you in production. To disable AWS/S3, set 'disable_aws-s3' to true in your newrelic.yml file. For more information on this instrumentation, check out our blog.
How to Add Custom Instrumentation
We encourage contributions to this project and will provide whatever assistance we can to those wishing to develop instrumentation for other open source Ruby libraries.
When adding instrumentation to this gem, be sure and get familiar with the RPM Agent API and contact [email protected] with any questions.
There are several extension points in the agent you can take advantage of with this gem.
- Custom tracers which measure methods and give visibility to otherwise unmeasured libraries.
- Samplers which sample some value about once a minute.
- Dispatcher support for web request handlers which would otherwise be undetected. In order for the agent to turn on in 'auto' mode it needs to discover a web dispatcher, or be started manually.
- Framework support, for alternatives to Rails like Camping or Ramaze
Custom Tracers
Custom tracers for frameworks should be added to the lib/rpm_contrib/instrumentation
directory. These files are loaded at the time the Agent starts. They will not
be loaded if the Agent does not start up.
It is important that you wrap any instrumentation with the checks necessary to determine if the code being instrumented is loaded. You can't add code to the contrib gem that will break when run in any other context besides yours.
For details on how to define custom tracers, refer to the support documentation on adding custom tracers. You can also get detailed information on the API from the Agent method tracing rdocs, especially the add_method_tracer docs.
A good example can be found in lib/rpm_contrib/instrumentation/paperclip.rb
.
Samplers
You can add samplers which will record metrics approximately once a minute. Samplers are useful for capturing generic instrumentation for display in custom views.
Samplers should extend the
NewRelic::Agent::Sampler
class. They should be placed in the samplers
directory.
Refer to examples in the RPM agent to see how to get started.
Supporting New Dispatchers
If you want to add support for a new dispatcher which is not being recognized by
default by the RPM agent, add code to the rpm_contrib/detection
directory.
This code needs to define a module in the NewRelic::LocalEnvironment
class.
This module will be accessed at the time environment detection takes place, when
the agent is initialized.
This module should define the method discover_dispatcher
and return the name
of the dispatcher if detected, or defer to super. See
rpm_contrib/detection/camping.rb
for a good example.
Supporting New Frameworks
Supporting new frameworks can be pretty involved and generally involves both adding custom instrumentation as well as framework and dispatcher detection.
In addition it will be necessary to define a new control class with the same
name as the framework. This control class must go in new_relic/control
.
Refer to the camping example in this gem to see how this is done in general.
If you decide to tackle any new frameworks, contact [email protected] and we'll be happy to help you work through it.
Note on Patches/Pull Requests
- Fork the http://www.github.com/newrelic/rpm_contrib project.
- Add instrumentation files to
lib/rpm_contrib/instrumentation
. These files will be loaded when the RPM agent is initialized. - Add samplers to
lib/rpm_contrib/samplers
. These classes are installed automatically when the RPM agent is initialized. - Add tests.
- Update README.md
- Commit, do not mess with the Rakefile, version, or history. (if you want to have your own version, that is fine but bump version in a commit by itself I can ignore when I pull)
- Send me a pull request. Bonus points for topic branches.
Further Information
Refer to the Agent API Documentation at http://newrelic.github.com/rpm
See the support site faqs at http://support.newrelic.com/faqs for additional tips and documentation.
Contact [email protected] for help.
Copyright
Copyright (c) 2009-2010 New Relic. See LICENSE for details.