deterministic_sampling.rb

puts <<-MSG

In order to drop entire traces at a certain rate, you should use
Honeycomb::DeterministicSampler. This uses a hash function on the trace ID to
make a deterministic but well-distributed choice as to whether the event should
be dropped or not at the given rate.

Here, we monkey-patch the (normally autogenerated) trace ID in two separate
traces. As it so happens, some human-friendly names hash such that we get the
desired results:

  should_sample(2, 'drop') #=> false
  should_sample(2, 'send') #=> true

Every event in the first trace will be dropped (so its URL won't work), and
every event in the second trace will be sent (so its URL will work, and there
won't be any orphans).

MSG

module Sampler
  extend Honeycomb::DeterministicSampler

  def self.sample(fields)
    rate = 2

    if should_sample(rate, fields['trace.trace_id'])
      puts "Send event: #{JSON.pretty_generate(fields)}\n\n"
      [true, rate]
    else
      puts "Drop event: #{JSON.pretty_generate(fields)}\n\n"
      [false, rate]
    end
  end
end

Honeycomb.configure do |config|
  config.write_key = ENV['HONEYCOMB_WRITE_KEY']
  config.dataset = 'exsampling'

  config.sample_hook do |fields|
    Sampler.sample(fields)
  end
end

Honeycomb.start_span(name: 'root') do
  Honeycomb.current_trace.instance_variable_set(:@id, 'drop')
  puts trace_url
  puts

  sleep 1

  Honeycomb.start_span(name: 'child 1') do
    sleep 1

    Honeycomb.start_span(name: 'grandchild') do
      sleep 1
    end
  end

  Honeycomb.start_span(name: 'child 2') do
    sleep 1
  end
end

puts

Honeycomb.start_span(name: 'root') do
  Honeycomb.current_trace.instance_variable_set(:@id, 'send')
  puts trace_url
  puts

  sleep 1

  Honeycomb.start_span(name: 'child 1') do
    sleep 1

    Honeycomb.start_span(name: 'grandchild') do
      sleep 1
    end
  end

  Honeycomb.start_span(name: 'child 2') do
    sleep 1
  end
end