Class: Delayed::Worker

Inherits:
Object
  • Object
show all
Includes:
Logging
Defined in:
lib/delayed/worker.rb,
lib/delayed/worker/health_check.rb,
lib/delayed/worker/process_helper.rb,
lib/delayed/worker/null_health_check.rb,
lib/delayed/worker/consul_health_check.rb

Defined Under Namespace

Modules: ProcessHelper Classes: ConsulHealthCheck, HealthCheck, NullHealthCheck

Constant Summary collapse

SIGNALS = 
%i[INT TERM QUIT].freeze
LINUX_PAGE_SIZE =

linux w/ proc fs

(size = `getconf PAGESIZE`.to_i
size.positive? ? size : 4096)

Class Attribute Summary collapse

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Methods included from Logging

log_job, logger, #say

Constructor Details

#initialize(options = {}) ⇒ Worker

Returns a new instance of Worker.



57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
 
# File 'lib/delayed/worker.rb', line 57

def initialize(options = {})
  @exit = false
  @parent_pid = options[:parent_pid]
  @queue_name = options[:queue] ||= Settings.queue
  @min_priority = options[:min_priority]
  @max_priority = options[:max_priority]
  @max_job_count = options[:worker_max_job_count].to_i
  @max_memory_usage = options[:worker_max_memory_usage].to_i
  @memory_high_water = options[:worker_memory_high_water]&.to_i || (@max_memory_usage / 2)
  @work_queue = options.delete(:work_queue) || WorkQueue::InProcess.new
  @health_check_type = Settings.worker_health_check_type
  @health_check_config = Settings.worker_health_check_config
  @config = options
  @job_count = 0

  @signal_queue = []

  plugins.each(&:inject!)
end

Class Attribute Details

.on_max_failures ⇒ Object

Callback to fire when a delayed job fails max_attempts times. If this callback is defined, then the value of destroy_failed_jobs is ignored, and the job is destroyed if this block returns true.

This allows for destroying “uninteresting” failures, while keeping around interesting failures to be investigated later.

The block is called with args(job, last_exception)



36
37
38
 
# File 'lib/delayed/worker.rb', line 36

def on_max_failures
  @on_max_failures
end

Instance Attribute Details

#config ⇒ Object (readonly)

Returns the value of attribute config.



25
26
27
 
# File 'lib/delayed/worker.rb', line 25

def config
  @config
end

#max_priority ⇒ Object (readonly)

Returns the value of attribute max_priority.



25
26
27
 
# File 'lib/delayed/worker.rb', line 25

def max_priority
  @max_priority
end

#min_priority ⇒ Object (readonly)

Returns the value of attribute min_priority.



25
26
27
 
# File 'lib/delayed/worker.rb', line 25

def min_priority
  @min_priority
end

#queue_name ⇒ Object (readonly)

Returns the value of attribute queue_name.



25
26
27
 
# File 'lib/delayed/worker.rb', line 25

def queue_name
  @queue_name
end

#work_queue ⇒ Object (readonly)

Returns the value of attribute work_queue.



25
26
27
 
# File 'lib/delayed/worker.rb', line 25

def work_queue
  @work_queue
end

Class Method Details

.current_job ⇒ Object 



46
47
48
 
# File 'lib/delayed/worker.rb', line 46

def self.current_job
  Thread.current[:running_delayed_job]
end

.lifecycle ⇒ Object 



42
43
44
 
# File 'lib/delayed/worker.rb', line 42

def self.lifecycle
  @lifecycle ||= Delayed::Lifecycle.new
end

.running_job(job) ⇒ Object 



50
51
52
53
54
55
 
# File 'lib/delayed/worker.rb', line 50

def self.running_job(job)
  Thread.current[:running_delayed_job] = job
  yield
ensure
  Thread.current[:running_delayed_job] = nil
end

Instance Method Details

#cleaned? ⇒ Boolean

Returns:

  • (Boolean) 


167
168
169
 
# File 'lib/delayed/worker.rb', line 167

def cleaned?
  @cleaned
end

#cleanup! ⇒ Object 



157
158
159
160
161
162
163
164
165
 
# File 'lib/delayed/worker.rb', line 157

def cleanup!
  return if cleaned?

  health_check.stop
  work_queue.close
  Delayed::Job.clear_locks!(name)

  @cleaned = true
end

#configure_for_job(job) ⇒ Object

set up the session context information, so that it gets logged with the job log lines also set up a unique tmpdir, which will get removed at the end of the job.



277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
 
# File 'lib/delayed/worker.rb', line 277

def configure_for_job(job)
  previous_tmpdir = ENV.fetch("TMPDIR", nil)

  self.class.running_job(job) do
    dir = Dir.mktmpdir("job-#{job.id}-#{name.gsub(/[^\w.]/, ".")}-")
    begin
      ENV["TMPDIR"] = dir
      yield
    ensure
      FileUtils.remove_entry(dir, true)
    end
  end
ensure
  ENV["TMPDIR"] = previous_tmpdir
end

#exit? ⇒ Boolean

Returns:

  • (Boolean) 


85
86
87
 
# File 'lib/delayed/worker.rb', line 85

def exit?
  !!@exit || parent_exited?
end

#handle_failed_job(job, error) ⇒ Object 



265
266
267
268
269
 
# File 'lib/delayed/worker.rb', line 265

def handle_failed_job(job, error)
  job.last_error = "#{error.message}\n#{error.backtrace.join("\n")}"
  logger.error("Failed with #{error.class} [#{error.message}] (#{job.attempts} attempts)")
  job.reschedule(error)
end

#health_check ⇒ Object 



293
294
295
296
297
298
299
 
# File 'lib/delayed/worker.rb', line 293

def health_check
  @health_check ||= HealthCheck.build(
    type: @health_check_type,
    worker_name: name,
    config: @health_check_config
  )
end

#id ⇒ Object 



271
272
273
 
# File 'lib/delayed/worker.rb', line 271

def id
  Process.pid
end

#name ⇒ Object 



77
78
79
 
# File 'lib/delayed/worker.rb', line 77

def name
  @name ||= "#{Socket.gethostname rescue "X"}:#{id}"
end

#parent_exited? ⇒ Boolean

Returns:

  • (Boolean) 


89
90
91
 
# File 'lib/delayed/worker.rb', line 89

def parent_exited?
  @parent_pid && @parent_pid != Process.ppid
end

#perform(job) ⇒ Object 



208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
 
# File 'lib/delayed/worker.rb', line 208

def perform(job)
  begin
    count = 1
    raise Delayed::Backend::JobExpired, "job expired at #{job.expires_at}" if job.expired?

    self.class.lifecycle.run_callbacks(:perform, self, job) do
      self.process_name = "run:#{Settings.worker_procname_prefix}#{job.id}:#{job.name}"
      logger.info("Processing #{log_job(job, :long)}")
      runtime = Benchmark.realtime do
        if job.batch?
          # each job in the batch will have perform called on it, so we don't
          # need a timeout around this
          count = perform_batch(job)
        else
          job.invoke_job
        end
        job.destroy
      end
      logger.info("Completed #{log_job(job, :short)} #{format("%.0fms", (runtime * 1000))}")
    end
  rescue ::Delayed::RetriableError => e
    can_retry = job.attempts + 1 < job.inferred_max_attempts
    callback_type = can_retry ? :retry : :error
    self.class.lifecycle.run_callbacks(callback_type, self, job, e) do
      handle_failed_job(job, e)
    end
  rescue SystemExit => e
    # There wasn't really a failure here so no callbacks and whatnot needed,
    # still reschedule the job though.
    job.reschedule(e)
  rescue Exception => e # rubocop:disable Lint/RescueException
    if e.is_a?(NoMemoryError)
      GC.start # try and free up some memory before reporting the error
      logger.debug "Could not allocate memory (max is #{@max_memory_usage}), dying"
      @exit = true
    end
    self.class.lifecycle.run_callbacks(:error, self, job, e) do
      handle_failed_job(job, e)
    end
  end
  count
end

#perform_batch(parent_job) ⇒ Object 



251
252
253
254
255
256
257
258
259
260
261
262
263
 
# File 'lib/delayed/worker.rb', line 251

def perform_batch(parent_job)
  batch = parent_job.payload_object
  return unless batch.mode == :serial

  batch.jobs.each do |job|
    job.source = parent_job.source
    job.create_and_lock!(name)
    configure_for_job(job) do
      perform(job)
    end
  end
  batch.items.size
end

#process_name=(new_name) ⇒ Object 



81
82
83
 
# File 'lib/delayed/worker.rb', line 81

def process_name=(new_name)
  $0 = "delayed:#{new_name}"
end

#run ⇒ Object 



171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
 
# File 'lib/delayed/worker.rb', line 171

def run
  return if exit?

  self.class.lifecycle.run_callbacks(:loop, self) do
    self.process_name =
      "pop:#{Settings.worker_procname_prefix}#{@queue_name}:#{min_priority || 0}:#{max_priority || "max"}"
    job = self.class.lifecycle.run_callbacks(:pop, self) do
      work_queue.get_and_lock_next_available(name, config)
    end

    if job
      configure_for_job(job) do
        @job_count += perform(job)

        if @max_job_count.positive? && @job_count >= @max_job_count
          logger.debug "Max job count of #{@max_job_count} exceeded, dying"
          @exit = true
        end

        if @memory_high_water.positive?
          memory = sample_memory
          if memory > @memory_high_water
            logger.debug "Memory usage of #{memory} exceeds high water of #{@memory_high_water}, dying"
            @exit = true
          else
            logger.debug "Memory usage: #{memory}"
          end
        end
      end
    else
      self.process_name =
        "wait:#{Settings.worker_procname_prefix}#{@queue_name}:#{min_priority || 0}:#{max_priority || "max"}"
      sleep(Settings.sleep_delay + (rand * Settings.sleep_delay_stagger)) unless exit?
    end
  end
end

#sample_memory ⇒ Object

generic unix solution



312
313
314
315
 
# File 'lib/delayed/worker.rb', line 312

def sample_memory
  s = File.read("/proc/#{Process.pid}/statm").to_i rescue 0
  s * LINUX_PAGE_SIZE / 1024
end

#start ⇒ Object 



98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
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
 
# File 'lib/delayed/worker.rb', line 98

def start
  logger.info "Starting worker"
  begin
    Process.setrlimit(:DATA, @max_memory_usage, @max_memory_usage * 2) if @max_memory_usage.positive?
  rescue Errno::EINVAL
    # couldn't set an OS-level limit
  end
  self.process_name =
    "start:#{Settings.worker_procname_prefix}#{@queue_name}:#{min_priority || 0}:#{max_priority || "max"}"
  @self_pipe = IO.pipe
  work_queue.init

  work_thread = Thread.current
  SIGNALS.each do |sig|
    trap(sig) do
      @signal_queue << sig
      wake_up
    end
  end

  raise "Could not register health_check" unless health_check.start

  signal_processor = Thread.new do
    loop do
      @self_pipe[0].read(1)
      case @signal_queue.pop
      when :INT, :TERM
        @exit = true # get the main thread to bail early if it's waiting for a job
        work_thread.raise(SystemExit) # Force the main thread to bail out of the current job
        cleanup! # we're going to get SIGKILL'd in a moment, so clean up asap
        break
      when :QUIT
        @exit = true
      else
        logger.error "Unknown signal '#{sig}' received"
      end
    end
  end

  self.class.lifecycle.run_callbacks(:execute, self) do
    run until exit?
  end

  logger.info "Stopping worker"
rescue => e
  Rails.logger.fatal("Child process died: #{e.inspect}") rescue nil
  self.class.lifecycle.run_callbacks(:exceptional_exit, self, e) { nil }
ensure
  cleanup!

  if signal_processor
    signal_processor.kill
    signal_processor.join
  end

  @self_pipe&.each(&:close)
  @self_pipe = nil
end

#wake_up ⇒ Object 



93
94
95
96
 
# File 'lib/delayed/worker.rb', line 93

def wake_up
  @self_pipe[1].write_nonblock(".", exception: false)
  work_queue.wake_up
end