$ bundle add inner_performance

$ rails inner_performance:install:migrations
$ rails db:migrate

mount InnerPerformance::Engine, at: '/inner_performance'

InnerPerformance.configure do |config|
  config.sample_rates = {
    # 2% of all the requests will be stored and analyzed. I would
    # recommend keeping it low because otherwise your database will
    # get full of events quite fast. It will also probably impact
    # your app's performance. As an example: In a Spree shop with
    # approx. 170 requests per minute, keeping it at default 2%
    # provides me more than enough data to analyze and locate the
    # bottlenecks.
    'process_action.action_controller' => (Rails.env.production? ? 2 : 100),

    # 100% of all the jobs will be stored and analyzed.
    'perform.active_job' => 100
  }

  # For how long events should be stored. All the ones that exceeds
  # this limit will be cleared out by InnerPerformance::CleanupJob.
  config.events_retention = 1.week

  # Used mostly by UI. Tells the UI what's the middle range when it
  # comes to event duration. For the example below, all the durations
  # between 0 and 200ms will be considered fast, all the ones between
  # 200ms and 999ms will be considered medium and all above 999ms will
  # be considered slow.
  config.medium_duration_range = [200, 999]

  # Rules for ignoring an event. There are two rules applied by default:
  # * sample_rates - operates on configured sample rate and drops events
  #   which do not got lucky when drawing a random number
  # * InnerPerformance job namespace - we don't want to save events for
  #   the job that saves the events because that leeds to infinite loop.
  #   Better not remove this rule as it will lead to stack overflow.
  config.ignore_rules.push(
    proc { |event| !event.is_a?(ActiveSupport::Notifications::Event) }
  )
end

InnerPerformance::CleanupJob.perform_later