Big Ruby Conf Bonus tracks

A whirlwind tour of Tony Arcieri's Celluloid and Eric Lindvall's Metriks

Transcript

1. You should know Celluloid and Metriks
   Adam Keys
   http://therealadam.com
   @therealadam
   March 2013
   

   View Slide

2. Concurrent objects in Ruby
   http://celluloid.io
   

   View Slide

3. What’s an actor model?
   Actor "daryl"
   Object ralph
   ...
   ...
   frob
   frob
   grok
   Thread 1
   Actor "larry"
   Object B
   ...
   frob
   frob
   grok
   grok
   Thread 2
   

   View Slide

4. Celluloid actors
   require 'redis'
   require 'celluloid'
   class RedisAdapter
   include Celluloid
   def initialize(redis)
   @redis = redis
   end
   def record(path)
   redis.incr("stats:#{path}")
   end
   def fetch(path)
   redis.get("stats:#{path}")
   end
   protected
   attr_reader :redis
   end
   

   View Slide

5. Celluloid actors
   require 'celluloid'
   require 'json'
   class Stats
   include Celluloid
   def handle(msg)
   log msg
   case msg
   when %r{^VISIT (.*)$}
   path = $1 # regex globals :(
   redis.async.record(path)
   "OK\n"
   when %r{^STATS (.*)$}
   path = $1 # regex globals :(
   stats = redis.fetch(path)
   JSON.dump(stats: {path => stats})
   else
   "ERR\n"
   end
   end
   def log(msg)
   Celluloid.logger.info(msg)
   end
   def redis
   Celluloid::Actor[:redis]
   end
   end
   

   View Slide

6. Reliability and Supervision Groups
   class Collector < Celluloid::SupervisionGroup
   supervise Stats, as: :stats
   supervise RedisAdapter, as: :redis, args: [Redis.new]
   supervise Listener, as: :listener, args: ["0.0.0.0", 4000]
   end
   if __FILE__ == $0
   collector = Collector.run!
   trap("INT") { collector.terminate; exit }
   sleep
   end
   

   View Slide

7. I/O actors
   require 'celluloid/io'
   class Listener
   include Celluloid::IO
   def initialize(host, port)
   log "Starting on #{host}:#{port}"
   @server = TCPServer.new(host, port)
   async.run
   end
   def finalize
   return unless @server
   @server.close
   end
   def run
   loop {
   async.handle_connection(@server.accept)
   }
   end
   def handle_connection(socket)
   _, host, port = socket.peeraddr
   log "#{host}: #{port} connected"
   msg = socket.readpartial(4096)
   response = stats.handle(msg)
   socket.write(response)
   socket.close
   rescue EOFError
   log "#{host}: #{port} disconnected"
   socket.close
   end
   def log(msg)
   puts msg
   end
   def stats
   Celluloid::Actor[:stats]
   end
   end
   

   View Slide

8. More reading
   •Erlang - http://erlang.org
   •D-Cell - https://github.com/celluloid/dcell
   •On Ruby concurrency:
   •http://therealadam.com/2012/06/19/getting-
   started-with-ruby-concurrency-using-two-
   simple-classes/
   •http://therealadam.com/2012/07/05/protect-
   that-state-locks-monitors-and-atomics/
   •http://therealadam.com/2012/09/10/designing-
   for-concurrency/
   •Java Concurrency in Practice
   

   View Slide

9. Knowing your production systems
   https://github.com/eric/metriks
   

   View Slide

10. All arguments are shallow given numbers
    

    View Slide

11. Counters, stats, et. al.
    require 'metrics'
    $api_hits = Metriks.counter('api_hits')
    $api_hits.increment
    require 'metriks/reporter/logger'
    reporter = Metriks::Reporter::Logger(logger: Logger.new('log/metrics.log'))
    reporter.start
    $api_stats = Metriks.timer('api_stats')
    $api_stats.time do
    # … work …
    end
    

    View Slide

12. 1. Logfiles
    I, [2013-02-18T07:00:00.006093 #50406] INFO -- : metriks: time=1361192400 name=stats type=timer count=92440
    one_minute_rate=434.4299373735234 five_minute_rate=
    220.682764284654 fifteen_minute_rate=90.15828364936169 mean_rate=423.37823163275544 min=0.000362023 max=0.070315041
    mean=0.0007417273702509757 stddev=2.4841137753254027e-06 median=0.000614221 95th_percentile=0.0011458962499999996
    I, [2013-02-18T07:00:00.009590 #50406] INFO -- : metriks: time=1361192400 name=api_stats type=timer count=92440
    one_minute_rate=434.4299373735234 five_minute_rate=220.682764284654 fifteen_minute_rate=90.15828364936169
    mean_rate=423.3715240226875 min=0.000300078 max=0.070015888 mean=0.0006167874175681445 stddev=2.4013694294628166e-06
    median=0.000505336 95th_percentile=0.0009915237499999996
    

    View Slide

13. 2. Convert to CSV
    #!/usr/bin/env ruby
    require 'csv'
    columns = %w{name time mean_rate 95th_percentile}
    out = CSV.generate do |csv|
    csv << columns.to_a
    ARGF.each_line do |l|
    fields = l.split
    next if fields[0] != 'I,'
    metrics = fields.slice(7..-1).inject({}) do |hsh, str|
    name, value = str.split('=')
    hsh.update(name => value)
    end
    csv << columns.inject([]) { |row, name| row << metrics[name] }
    end
    end
    puts out
    

    View Slide

14. 3. Spreadsheets!
    

    View Slide

15. Dashboard services
    

    View Slide

16. More reading
    •Metrics, the original library - http://
    metrics.codahale.com
    •Metrics everywhere, the slides - http://
    codahale.com/codeconf-2011-04-09-metrics-
    metrics-everywhere.pdf
    •Metrics everywhere, the talk - http://
    www.youtube.com/watch?v=czes-oa0yik
    

    View Slide

17. Thanks!
    

    View Slide