InstStatsd

configurable statsd client proxy

Set a few enviroment variables:

export INST_STATSD_HOST=statsd.example.org
export INST_STATSD_PORT=1234
export INST_STATSD_NAMESPACE=my_app.prod
export INST_STATSD_APPEND_HOSTNAME=false
export INST_DOG_TAGS='{"app": "canvas", "env": "prod"}'

Or pass a hash to InstStatsd.settings

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

InstStatsd.settings = settings

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

For statsd only the host (or INST_STATSD_HOST ENV var) is required, all other config are optional.

For data dog, only the dog_tags (or INST_DOG_TAGS ENV vars) are required. All others are optional. Having dog_tags is how to know to use data_dog. A {} can be passed if you have no desired default tags. The string needs to be parsable as json.

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

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

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'
}

InstStatsd.settings = settings

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

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

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)

These are the default tags for the application. Since we do not use the host for tags, this allows us to separate different apps and view metrics for one app. The tags are a hash. Example {app: canvas, environment: production} Data dog charges metrics based on unique combination of tags.

Outside of configuration, app code generally interacts with the InstStatsd::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.. }
InstStatsd::Statsd.timing("my_stat", ms)

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

InstStatsd 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/inst_statsd.rb
InstStatsd::DefaultTracking.track_sql
InstStatsd::DefaultTracking.track_cache
InstStatsd::DefaultTracking.track_active_record
InstStatsd::RequestTracking.enable

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

* as reported by ActiveSupport::Notifications

** as reported by aroi

If you'd like InstStatsd 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
InstStatsd::RequestTracking.enable logger: Rails.logger

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.

InstStatsd::DefaultTracking.track_sql
InstStatsd::DefaultTracking.track_cache
InstStatsd::DefaultTracking.track_active_record
InstStatsd::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 InstStatsd which category to track along:

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

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