CanvasStatsd

configurable statsd client proxy

Configuration

Set a few enviroment variables:

export CANVAS_STATSD_HOST=statsd.example.org
export CANVAS_STATSD_PORT=1234
export CANVAS_STATSD_NAMESPACE=my_app.prod
export CANVAS_STATSD_APPEND_HOSTNAME=false

Or pass a hash to CanvasStatsd.settings

settings = {
  host: 'statsd.example.org'
  port: 1234
  namespace: 'my_app.prod'
  append_hostname: false
}

CanvasStatsd.settings = settings

Values passed to CanvasStatsd.settings will be merged into and take precedence over any existing ENV vars

Configuration Options

Only the host (or CANVAS_STATSD_HOST ENV var) is required, all other config is optional

host

Location of the statsd box you want to send stats to.

port

port of the statsd box you want to send stats to.

namespace

If a namespace is defined, it'll be prepended to the stat name. So the following:

settings = {
  host: 'statsd.example.org'
  namespace: 'my_app.prod'
}

CanvasStatsd.settings = settings

CanvasStatsd::Statsd.timing('some.stat', 300)

would use my_app.prod.some.stat as it's stat name.

append_hostname

The hostname of the server will be appended to the stat name, unless append_hostname: false is specified. So if the namespace is canvas and the hostname is app01, the final stat name of my_stat would be canvas.my_stat.app01 (assuming the default statsd/graphite configuration)

Usage

Outside of configuration, app code generally interacts with the CanvasStatsd::Statsd object, which is a proxy class to communicate messages to statsd.

Available statsd messages are described in:

• etsty/statsd README
• reinh/statsd source

So for instance:

ms = Benchmark.ms { ..code.. }
CanvasStatsd::Statsd.timing("my_stat", ms)

If statsd isn't configured and enabled, then calls to CanvasStatsd::Statsd.* will do nothing and return nil.

Default Metrics Tracking

CanvasStatsd ships with a several trackers that can capture several performance metrics. To enable these default metrics tracking in your rails app, you enable the ones you want, and then enable request tracking:

# config/initializers/canvas_statsd.rb
CanvasStatsd::DefaultTracking.track_sql
CanvasStatsd::DefaultTracking.track_cache
CanvasStatsd::DefaultTracking.track_active_record
CanvasStatsd::RequestTracking.enable

This will track the following (as statsd timings) per request:

Metric Type	Statsd key	Description
total	controller.action.total	total time spent on controller action*
db	controller.action.db	time spent in the db*
view	controller.action.view	time spent build views*
sql write	controller.action.sql.write	number of sql writes
sql read	controller.action.sql.read	number of sql reads
sql cache	controller.action.sql.cache	number of sql cache
active record	controller.action.active_record	number of ActiveRecord objects created **
cache read	controller.action.cache.read	number of cache reads

* as reported by ActiveSupport::Notifications

** as reported by aroi

If you'd like CanvasStatsd to log these metrics (as well as sending them to statsd), pass a logger object along like so:

# log default metrics to environment logs in Rails
CanvasStatsd::RequestTracking.enable logger: Rails.logger

Block tracking

You can easily track the performance of any block of code using all enabled metrics. Just be careful that your key isn't too dynamic, causing performance problems for your statsd server.

CanvasStatsd::DefaultTracking.track_sql
CanvasStatsd::DefaultTracking.track_cache
CanvasStatsd::DefaultTracking.track_active_record
CanvasStatsd::BlockTracking.track("my_important_job") do
  sleep(10)
end

If you want to keep track of both exclusive and inclusive times for a re-entrant piece of code, you just need to tell CanvasStatsd which category to track along:

CanvasStatsd::BlockTracking.track("my_important_job", category: :my_stuff) do
  sleep(10)
  CanvasStatsd::BlockTracking.track("my_other_important_job", category: :my_stuff) do
    sleep(5)
  end
end

Contributing

1. Fork it
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request