Class: Resque::Worker

Inherits:
Object
  • Object
show all
Extended by:
Helpers
Includes:
Helpers
Defined in:
lib/resque/worker.rb

Overview

A Resque Worker processes jobs. On platforms that support fork(2), the worker will fork off a child to process each job. This ensures a clean slate when beginning the next job and cuts down on gradual memory growth as well as low level failures.

It also ensures workers are always listening to signals from you, their master, and can react accordingly.

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Methods included from Helpers

classify, constantize, decode, encode, redis

Constructor Details

#initialize(*queues) ⇒ Worker

Workers should be initialized with an array of string queue names. The order is important: a Worker will check the first queue given for a job. If none is found, it will check the second queue name given. If a job is found, it will be processed. Upon completion, the Worker will again check the first queue given, and so forth. In this way the queue list passed to a Worker on startup defines the priorities of queues.

If passed a single “*”, this Worker will operate on all queues in alphabetical order. Queues can be dynamically added or removed without needing to restart workers using this method.



95
96
97
98
99
100
 
# File 'lib/resque/worker.rb', line 95

def initialize(*queues)
  @queues = queues.map { |queue| queue.to_s.strip }
  @shutdown = nil
  @paused = nil
  validate_queues
end

Instance Attribute Details

#cant_fork ⇒ Object

Boolean indicating whether this worker can or can not fork. Automatically set if a fork(2) fails.



21
22
23
 
# File 'lib/resque/worker.rb', line 21

def cant_fork
  @cant_fork
end

#term_child ⇒ Object

decide whether to use new_kill_child logic



26
27
28
 
# File 'lib/resque/worker.rb', line 26

def term_child
  @term_child
end

#term_timeout ⇒ Object

Returns the value of attribute term_timeout.



23
24
25
 
# File 'lib/resque/worker.rb', line 23

def term_timeout
  @term_timeout
end

#to_s ⇒ Object Also known as: id

The string representation is the same as the id for this worker instance. Can be used with ‘Worker.find`.



552
553
554
 
# File 'lib/resque/worker.rb', line 552

def to_s
  @to_s ||= "#{hostname}:#{Process.pid}:#{@queues.join(',')}"
end

#verbose ⇒ Object

Whether the worker should log basic info to STDOUT



14
15
16
 
# File 'lib/resque/worker.rb', line 14

def verbose
  @verbose
end

#very_verbose ⇒ Object

Whether the worker should log lots of info to STDOUT



17
18
19
 
# File 'lib/resque/worker.rb', line 17

def very_verbose
  @very_verbose
end

Class Method Details

.all ⇒ Object

Returns an array of all worker objects.



31
32
33
 
# File 'lib/resque/worker.rb', line 31

def self.all
  Array(redis.smembers(:workers)).map { |id| find(id) }.compact
end

.attach(worker_id) ⇒ Object

Alias of ‘find`



74
75
76
 
# File 'lib/resque/worker.rb', line 74

def self.attach(worker_id)
  find(worker_id)
end

.exists?(worker_id) ⇒ Boolean

Given a string worker id, return a boolean indicating whether the worker exists

Returns:

  • (Boolean) 


80
81
82
 
# File 'lib/resque/worker.rb', line 80

def self.exists?(worker_id)
  redis.sismember(:workers, worker_id)
end

.find(worker_id) ⇒ Object

Returns a single worker object. Accepts a string id.



62
63
64
65
66
67
68
69
70
71
 
# File 'lib/resque/worker.rb', line 62

def self.find(worker_id)
  if exists? worker_id
    queues = worker_id.split(':')[-1].split(',')
    worker = new(*queues)
    worker.to_s = worker_id
    worker
  else
    nil
  end
end

.working ⇒ Object

Returns an array of all worker objects currently processing jobs.



37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
 
# File 'lib/resque/worker.rb', line 37

def self.working
  names = all
  return [] unless names.any?

  names.map! { |name| "worker:#{name}" }

  reportedly_working = {}

  begin
    reportedly_working = redis.mapped_mget(*names).reject do |key, value|
      value.nil? || value.empty?
    end
  rescue Redis::Distributed::CannotDistribute
    names.each do |name|
      value = redis.get name
      reportedly_working[name] = value unless value.nil? || value.empty?
    end
  end

  reportedly_working.keys.map do |key|
    find key.sub("worker:", '')
  end.compact
end

Instance Method Details

#==(other) ⇒ Object

Is this worker the same as another worker?



542
543
544
 
# File 'lib/resque/worker.rb', line 542

def ==(other)
  to_s == other.to_s
end

#done_working ⇒ Object

Called when we are done working - clears our ‘working_on` state and tells Redis we processed a job.



478
479
480
481
 
# File 'lib/resque/worker.rb', line 478

def done_working
  processed!
  redis.del("worker:#{self}")
end

#enable_gc_optimizations ⇒ Object

Enables GC Optimizations if you’re running REE. www.rubyenterpriseedition.com/faq.html#adapt_apps_for_cow



285
286
287
288
289
 
# File 'lib/resque/worker.rb', line 285

def enable_gc_optimizations
  if GC.respond_to?(:copy_on_write_friendly=)
    GC.copy_on_write_friendly = true
  end
end

#failed ⇒ Object

How many failed jobs has this worker seen? Returns an int.



495
496
497
 
# File 'lib/resque/worker.rb', line 495

def failed
  Stat["failed:#{self}"]
end

#failed! ⇒ Object

Tells Redis we’ve failed a job.



500
501
502
503
 
# File 'lib/resque/worker.rb', line 500

def failed!
  Stat << "failed"
  Stat << "failed:#{self}"
end

#fork(job) ⇒ Object

Not every platform supports fork. Here we do our magic to determine if yours does.



249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
 
# File 'lib/resque/worker.rb', line 249

def fork(job)
  return if @cant_fork
  
  # Only run before_fork hooks if we're actually going to fork
  # (after checking @cant_fork)
  run_hook :before_fork, job

  begin
    # IronRuby doesn't support `Kernel.fork` yet
    if Kernel.respond_to?(:fork)
      Kernel.fork if will_fork?
    else
      raise NotImplementedError
    end
  rescue NotImplementedError
    @cant_fork = true
    nil
  end
end

#hostname ⇒ Object

chomp’d hostname of this machine



558
559
560
 
# File 'lib/resque/worker.rb', line 558

def hostname
  @hostname ||= `hostname`.chomp
end

#idle? ⇒ Boolean

Boolean - true if idle, false if not

Returns:

  • (Boolean) 


527
528
529
 
# File 'lib/resque/worker.rb', line 527

def idle?
  state == :idle
end

#inspect ⇒ Object 



546
547
548
 
# File 'lib/resque/worker.rb', line 546

def inspect
  "#<Worker #{to_s}>"
end

#job ⇒ Object Also known as: processing

Returns a hash explaining the Job we’re currently processing, if any.



516
517
518
 
# File 'lib/resque/worker.rb', line 516

def job
  decode(redis.get("worker:#{self}")) || {}
end

#kill_child ⇒ Object

Kills the forked child immediately, without remorse. The job it is processing will not be completed.



355
356
357
358
359
360
361
362
363
364
365
 
# File 'lib/resque/worker.rb', line 355

def kill_child
  if @child
    log! "Killing child at #{@child}"
    if system("ps -o pid,state -p #{@child}")
      Process.kill("KILL", @child) rescue nil
    else
      log! "Child #{@child} not found, restarting."
      shutdown
    end
  end
end

#linux_worker_pids ⇒ Object

Find Resque worker pids on Linux and OS X.

Returns an Array of string pids of all the other workers on this machine. Useful when pruning dead workers on startup.



581
582
583
584
585
 
# File 'lib/resque/worker.rb', line 581

def linux_worker_pids
  `ps -A -o pid,command | grep "[r]esque" | grep -v "resque-web"`.split("\n").map do |line|
    line.split(' ')[0]
  end
end

#log(message) ⇒ Object

Log a message to STDOUT if we are verbose or very_verbose.



610
611
612
613
614
615
616
617
 
# File 'lib/resque/worker.rb', line 610

def log(message)
  if verbose
    puts "*** #{message}"
  elsif very_verbose
    time = Time.now.strftime('%H:%M:%S %Y-%m-%d')
    puts "** [#{time}] #$$: #{message}"
  end
end

#log!(message) ⇒ Object

Logs a very verbose message to STDOUT.



620
621
622
 
# File 'lib/resque/worker.rb', line 620

def log!(message)
  log message if very_verbose
end

#new_kill_child ⇒ Object

Kills the forked child immediately with minimal remorse. The job it is processing will not be completed. Send the child a TERM signal, wait 5 seconds, and then a KILL signal if it has not quit



370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
 
# File 'lib/resque/worker.rb', line 370

def new_kill_child
  if @child
    unless Process.waitpid(@child, Process::WNOHANG)
      log! "Sending TERM signal to child #{@child}"
      Process.kill("TERM", @child)
      (term_timeout.to_f * 10).round.times do |i|
        sleep(0.1)
        return if Process.waitpid(@child, Process::WNOHANG)
      end
      log! "Sending KILL signal to child #{@child}"
      Process.kill("KILL", @child)
    else
      log! "Child #{@child} already quit."
    end
  end
rescue SystemCallError
  log! "Child #{@child} already quit and reaped."
end

#pause_processing ⇒ Object

Stop processing jobs after the current one has completed (if we’re currently running one).



396
397
398
399
 
# File 'lib/resque/worker.rb', line 396

def pause_processing
  log "USR2 received; pausing job processing"
  @paused = true
end

#paused? ⇒ Boolean

are we paused?

Returns:

  • (Boolean) 


390
391
392
 
# File 'lib/resque/worker.rb', line 390

def paused?
  @paused
end

#perform(job) ⇒ Object

Processes a given job in the child.



185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
 
# File 'lib/resque/worker.rb', line 185

def perform(job)
  begin
    run_hook :after_fork, job if will_fork?
    job.perform
  rescue Object => e
    log "#{job.inspect} failed: #{e.inspect}"
    begin
      job.fail(e)
    rescue Object => e
      log "Received exception when reporting failure: #{e.inspect}"
    end
    failed!
  else
    log "done: #{job.inspect}"
  ensure
    yield job if block_given?
  end
end

#pid ⇒ Object

Returns Integer PID of running worker



563
564
565
 
# File 'lib/resque/worker.rb', line 563

def pid
  Process.pid
end

#process(job = nil, &block) ⇒ Object

DEPRECATED. Processes a single job. If none is given, it will try to produce one. Usually run in the child.



174
175
176
177
178
179
180
181
182
 
# File 'lib/resque/worker.rb', line 174

def process(job = nil, &block)
  return unless job ||= reserve

  job.worker = self
  working_on job
  perform(job, &block)
ensure
  done_working
end

#processed ⇒ Object

How many jobs has this worker processed? Returns an int.



484
485
486
 
# File 'lib/resque/worker.rb', line 484

def processed
  Stat["processed:#{self}"]
end

#processed! ⇒ Object

Tell Redis we’ve processed a job.



489
490
491
492
 
# File 'lib/resque/worker.rb', line 489

def processed!
  Stat << "processed"
  Stat << "processed:#{self}"
end

#procline(string) ⇒ Object

Given a string, sets the procline ($0) and logs. Procline is always in the format of:

resque-VERSION: STRING



604
605
606
607
 
# File 'lib/resque/worker.rb', line 604

def procline(string)
  $0 = "resque-#{Resque::Version}: #{string}"
  log! $0
end

#prune_dead_workers ⇒ Object

Looks for any workers which should be running on this server and, if they’re not, removes them from Redis.

This is a form of garbage collection. If a server is killed by a hard shutdown, power failure, or something else beyond our control, the Resque workers will not die gracefully and therefore will leave stale state information in Redis.

By checking the current Redis state against the actual environment, we can determine if Redis is old and clean it up a bit.



417
418
419
420
421
422
423
424
425
426
427
 
# File 'lib/resque/worker.rb', line 417

def prune_dead_workers
  all_workers = Worker.all
  known_workers = worker_pids unless all_workers.empty?
  all_workers.each do |worker|
    host, pid, queues = worker.id.split(':')
    next unless host == hostname
    next if known_workers.include?(pid)
    log! "Pruning dead worker: #{worker}"
    worker.unregister_worker
  end
end

#queues ⇒ Object

Returns a list of queues to use when searching for a job. A splat (“*”) means you want every queue (in alpha order) - this can be useful for dynamically adding new queues.



243
244
245
 
# File 'lib/resque/worker.rb', line 243

def queues
  @queues.map {|queue| queue == "*" ? Resque.queues.sort : queue }.flatten.uniq
end

#reconnect ⇒ Object

Reconnect to Redis to avoid sharing a connection with the parent, retry up to 3 times with increasing delay before giving up.



224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
 
# File 'lib/resque/worker.rb', line 224

def reconnect
  tries = 0
  begin
    redis.client.reconnect
  rescue Redis::BaseConnectionError
    if (tries += 1) <= 3
      log "Error reconnecting to Redis; retrying"
      sleep(tries)
      retry
    else
      log "Error reconnecting to Redis; quitting"
      raise
    end
  end
end

#register_signal_handlers ⇒ Object

Registers the various signal handlers a worker responds to.

TERM: Shutdown immediately, stop processing jobs.

INT: Shutdown immediately, stop processing jobs.

QUIT: Shutdown after the current job has finished processing. USR1: Kill the forked child immediately, continue processing jobs. USR2: Don’t process any new jobs CONT: Start processing jobs again after a USR2



299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
 
# File 'lib/resque/worker.rb', line 299

def register_signal_handlers
  trap('TERM') { shutdown!  }
  trap('INT')  { shutdown!  }

  begin
    trap('QUIT') { shutdown   }
    if term_child
      trap('USR1') { new_kill_child }
    else
      trap('USR1') { kill_child }
    end
    trap('USR2') { pause_processing }
    trap('CONT') { unpause_processing }
  rescue ArgumentError
    warn "Signals QUIT, USR1, USR2, and/or CONT not supported."
  end

  log! "Registered signals"
end

#register_worker ⇒ Object

Registers ourself as a worker. Useful when entering the worker lifecycle on startup.



431
432
433
434
 
# File 'lib/resque/worker.rb', line 431

def register_worker
  redis.sadd(:workers, self)
  started!
end

#reserve ⇒ Object

Attempts to grab a job off one of the provided queues. Returns nil if no job can be found.



206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
 
# File 'lib/resque/worker.rb', line 206

def reserve
  queues.each do |queue|
    log! "Checking #{queue}"
    if job = Resque.reserve(queue)
      log! "Found job on #{queue}"
      return job
    end
  end

  nil
rescue Exception => e
  log "Error reserving job: #{e.inspect}"
  log e.backtrace.join("\n")
  raise e
end

#run_hook(name, *args) ⇒ Object

Runs a named hook, passing along any arguments.



437
438
439
440
441
442
443
444
 
# File 'lib/resque/worker.rb', line 437

def run_hook(name, *args)
  return unless hook = Resque.send(name)
  msg = "Running #{name} hook"
  msg << " with #{args.inspect}" if args.any?
  log msg

  args.any? ? hook.call(*args) : hook.call
end

#shutdown ⇒ Object

Schedule this worker for shutdown. Will finish processing the current job.



333
334
335
336
 
# File 'lib/resque/worker.rb', line 333

def shutdown
  log 'Exiting...'
  @shutdown = true
end

#shutdown! ⇒ Object

Kill the child and shutdown immediately.



339
340
341
342
343
344
345
346
 
# File 'lib/resque/worker.rb', line 339

def shutdown!
  shutdown
  if term_child
    new_kill_child
  else
    kill_child
  end
end

#shutdown? ⇒ Boolean

Should this worker shutdown as soon as current job is finished?

Returns:

  • (Boolean) 


349
350
351
 
# File 'lib/resque/worker.rb', line 349

def shutdown?
  @shutdown
end

#solaris_worker_pids ⇒ Object

Find Resque worker pids on Solaris.

Returns an Array of string pids of all the other workers on this machine. Useful when pruning dead workers on startup.



591
592
593
594
595
596
597
598
599
 
# File 'lib/resque/worker.rb', line 591

def solaris_worker_pids
  `ps -A -o pid,comm | grep "[r]uby" | grep -v "resque-web"`.split("\n").map do |line|
    real_pid = line.split(' ')[0]
    pargs_command = `pargs -a #{real_pid} 2>/dev/null | grep [r]esque | grep -v "resque-web"`
    if pargs_command.split(':')[1] == " resque-#{Resque::Version}"
      real_pid
    end
  end.compact
end

#started ⇒ Object

What time did this worker start? Returns an instance of ‘Time`



506
507
508
 
# File 'lib/resque/worker.rb', line 506

def started
  redis.get "worker:#{self}:started"
end

#started! ⇒ Object

Tell Redis we’ve started



511
512
513
 
# File 'lib/resque/worker.rb', line 511

def started!
  redis.set("worker:#{self}:started", Time.now.to_s)
end

#startup ⇒ Object

Runs all the methods needed when a worker begins its lifecycle.



270
271
272
273
274
275
276
277
278
279
280
281
 
# File 'lib/resque/worker.rb', line 270

def startup
  warn "WARNING: This way of doing signal handling is now deprecated. Please see http://hone.heroku.com/resque/2012/08/21/resque-signals.html for more info." unless term_child
  enable_gc_optimizations
  register_signal_handlers
  prune_dead_workers
  run_hook :before_first_fork
  register_worker

  # Fix buffering so we can `rake resque:work > resque.log` and
  # get output from the child in there.
  $stdout.sync = true
end

#state ⇒ Object

Returns a symbol representing the current worker state, which can be either :working or :idle



537
538
539
 
# File 'lib/resque/worker.rb', line 537

def state
  redis.exists("worker:#{self}") ? :working : :idle
end

#unpause_processing ⇒ Object

Start processing jobs again after a pause



402
403
404
405
 
# File 'lib/resque/worker.rb', line 402

def unpause_processing
  log "CONT received; resuming job processing"
  @paused = false
end

#unregister_signal_handlers ⇒ Object 



319
320
321
322
323
324
325
326
327
328
329
 
# File 'lib/resque/worker.rb', line 319

def unregister_signal_handlers
  trap('TERM') { raise TermException.new("SIGTERM") }
  trap('INT', 'DEFAULT')

  begin
    trap('QUIT', 'DEFAULT')
    trap('USR1', 'DEFAULT')
    trap('USR2', 'DEFAULT')
  rescue ArgumentError
  end
end

#unregister_worker(exception = nil) ⇒ Object

Unregisters ourself as a worker. Useful when shutting down.



447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
 
# File 'lib/resque/worker.rb', line 447

def unregister_worker(exception = nil)
  # If we're still processing a job, make sure it gets logged as a
  # failure.
  if (hash = processing) && !hash.empty?
    job = Job.new(hash['queue'], hash['payload'])
    # Ensure the proper worker is attached to this job, even if
    # it's not the precise instance that died.
    job.worker = self
    job.fail(exception || DirtyExit.new)
  end

  redis.srem(:workers, self)
  redis.del("worker:#{self}")
  redis.del("worker:#{self}:started")

  Stat.clear("processed:#{self}")
  Stat.clear("failed:#{self}")
end

#validate_queues ⇒ Object

A worker must be given a queue, otherwise it won’t know what to do with itself.

You probably never need to call this.



106
107
108
109
110
 
# File 'lib/resque/worker.rb', line 106

def validate_queues
  if @queues.nil? || @queues.empty?
    raise NoQueueError.new("Please give each worker at least one queue.")
  end
end

#will_fork? ⇒ Boolean

Returns:

  • (Boolean) 


531
532
533
 
# File 'lib/resque/worker.rb', line 531

def will_fork?
  !(@cant_fork || $TESTING)
end

#work(interval = 5.0, &block) ⇒ Object

This is the main workhorse method. Called on a Worker instance, it begins the worker life cycle.

The following events occur during a worker’s life cycle:

  1. Startup: Signals are registered, dead workers are pruned,

    and this worker is registered.

  2. Work loop: Jobs are pulled from a queue and processed.

  3. Teardown: This worker is unregistered.

Can be passed a float representing the polling frequency. The default is 5 seconds, but for a semi-active site you may want to use a smaller value.

Also accepts a block which will be passed the job as soon as it has completed processing. Useful for testing.



128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
 
# File 'lib/resque/worker.rb', line 128

def work(interval = 5.0, &block)
  interval = Float(interval)
  $0 = "resque: Starting"
  startup

  loop do
    break if shutdown?

    if not paused? and job = reserve
      log "got: #{job.inspect}"
      job.worker = self
      working_on job

      if @child = fork(job)
        srand # Reseeding
        procline "Forked #{@child} at #{Time.now.to_i}"
        begin
          Process.waitpid(@child)
        rescue SystemCallError
          nil
        end
      else
        unregister_signal_handlers if will_fork? && term_child
        procline "Processing #{job.queue} since #{Time.now.to_i}"
        reconnect
        perform(job, &block)
        exit! if will_fork?
      end

      done_working
      @child = nil
    else
      break if interval.zero?
      log! "Sleeping for #{interval} seconds"
      procline paused? ? "Paused" : "Waiting for #{@queues.join(',')}"
      sleep interval
    end
  end

  unregister_worker
rescue Exception => exception
  unregister_worker(exception)
end

#worker_pids ⇒ Object

Returns an Array of string pids of all the other workers on this machine. Useful when pruning dead workers on startup.



569
570
571
572
573
574
575
 
# File 'lib/resque/worker.rb', line 569

def worker_pids
  if RUBY_PLATFORM =~ /solaris/
    solaris_worker_pids
  else
    linux_worker_pids
  end
end

#working? ⇒ Boolean

Boolean - true if working, false if not

Returns:

  • (Boolean) 


522
523
524
 
# File 'lib/resque/worker.rb', line 522

def working?
  state == :working
end

#working_on(job) ⇒ Object

Given a job, tells Redis we’re working on it. Useful for seeing what workers are doing and when.



468
469
470
471
472
473
474
 
# File 'lib/resque/worker.rb', line 468

def working_on(job)
  data = encode \
    :queue   => job.queue,
    :run_at  => Time.now.strftime("%Y/%m/%d %H:%M:%S %Z"),
    :payload => job.payload
  redis.set("worker:#{self}", data)
end