Class: Bluepill::Process

Inherits:
Object
  • Object
show all
Defined in:
lib/bluepill/process.rb

Constant Summary collapse

CONFIGURABLE_ATTRIBUTES = 
[
  :pre_start_command,
  :start_command,
  :stop_command,
  :restart_command,

  :stdout,
  :stderr,
  :stdin,

  :daemonize,
  :pid_file,
  :working_dir,
  :environment,

  :start_grace_time,
  :stop_grace_time,
  :restart_grace_time,

  :uid,
  :gid,

  :cache_actual_pid,

  :monitor_children,
  :child_process_factory,

  :pid_command,
  :auto_start,

  :supplementary_groups,

  :stop_signals,

  :on_start_timeout,

  :group_start_noblock,
  :group_restart_noblock,
  :group_stop_noblock,
  :group_unmonitor_noblock,

]

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(process_name, checks, options = {}) ⇒ Process

Returns a new instance of Process.



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
 
# File 'lib/bluepill/process.rb', line 112

def initialize(process_name, checks, options = {})
  @name = process_name
  @event_mutex = Monitor.new
  @watches = []
  @triggers = []
  @children = []
  @threads = []
  @statistics = ProcessStatistics.new
  @actual_pid = options[:actual_pid]
  self.logger = options[:logger]

  checks.each do |name, opts|
    if Trigger[name]
      add_trigger(name, opts)
    else
      add_watch(name, opts)
    end
  end

  # These defaults are overriden below if it's configured to be something else.
  @monitor_children = false
  @cache_actual_pid = true
  @start_grace_time = @stop_grace_time = @restart_grace_time = 3
  @environment = {}
  @on_start_timeout = 'start'
  @group_start_noblock = @group_stop_noblock = @group_restart_noblock = @group_unmonitor_noblock = true

  CONFIGURABLE_ATTRIBUTES.each do |attribute_name|
    send("#{attribute_name}=", options[attribute_name]) if options.key?(attribute_name)
  end

  # Let state_machine do its initialization stuff
  super() # no arguments intentional
end

Instance Attribute Details

#children ⇒ Object (readonly)

Returns the value of attribute children.



56
57
58
 
# File 'lib/bluepill/process.rb', line 56

def children
  @children
end

#logger ⇒ Object

Returns the value of attribute logger.



54
55
56
 
# File 'lib/bluepill/process.rb', line 54

def logger
  @logger
end

#name ⇒ Object

Returns the value of attribute name.



54
55
56
 
# File 'lib/bluepill/process.rb', line 54

def name
  @name
end

#process_running ⇒ Object

Returns the value of attribute process_running.



54
55
56
 
# File 'lib/bluepill/process.rb', line 54

def process_running
  @process_running
end

#skip_ticks_until ⇒ Object

Returns the value of attribute skip_ticks_until.



54
55
56
 
# File 'lib/bluepill/process.rb', line 54

def skip_ticks_until
  @skip_ticks_until
end

#statistics ⇒ Object (readonly)

Returns the value of attribute statistics.



56
57
58
 
# File 'lib/bluepill/process.rb', line 56

def statistics
  @statistics
end

#triggers ⇒ Object

Returns the value of attribute triggers.



54
55
56
 
# File 'lib/bluepill/process.rb', line 54

def triggers
  @triggers
end

#watches ⇒ Object

Returns the value of attribute watches.



54
55
56
 
# File 'lib/bluepill/process.rb', line 54

def watches
  @watches
end

Instance Method Details

#actual_pid ⇒ Object 



423
424
425
 
# File 'lib/bluepill/process.rb', line 423

def actual_pid
  pid_command ? pid_from_command : pid_from_file
end

#actual_pid=(pid) ⇒ Object 



447
448
449
450
 
# File 'lib/bluepill/process.rb', line 447

def actual_pid=(pid)
  ProcessJournal.append_pid_to_journal(name, pid) # be sure to always log the pid
  @actual_pid = pid
end

#add_trigger(name, options = {}) ⇒ Object 



213
214
215
 
# File 'lib/bluepill/process.rb', line 213

def add_trigger(name, options = {})
  triggers << Trigger[name].new(self, options.merge(logger: logger))
end

#add_watch(name, options = {}) ⇒ Object

Watch related methods



209
210
211
 
# File 'lib/bluepill/process.rb', line 209

def add_watch(name, options = {})
  watches << ConditionWatch.new(name, options.merge(logger: logger))
end

#cache_actual_pid? ⇒ Boolean

Returns:

  • (Boolean) 


419
420
421
 
# File 'lib/bluepill/process.rb', line 419

def cache_actual_pid?
  !!@cache_actual_pid
end

#clean_threads ⇒ Object 



397
398
399
400
 
# File 'lib/bluepill/process.rb', line 397

def clean_threads
  @threads.each(&:kill)
  @threads.clear
end

#cleanup_process ⇒ Object 



392
393
394
395
 
# File 'lib/bluepill/process.rb', line 392

def cleanup_process
  ProcessJournal.kill_all_from_journal(name) # finish cleanup
  unlink_pid # TODO: we only write the pid file if we daemonize, should we only unlink it if we daemonize?
end

#clear_pid ⇒ Object 



452
453
454
 
# File 'lib/bluepill/process.rb', line 452

def clear_pid
  @actual_pid = nil
end

#daemonize? ⇒ Boolean

Returns:

  • (Boolean) 


402
403
404
 
# File 'lib/bluepill/process.rb', line 402

def daemonize?
  !!daemonize
end

#determine_initial_state ⇒ Object 



239
240
241
242
243
244
245
 
# File 'lib/bluepill/process.rb', line 239

def determine_initial_state
  if self.process_running?(true)
    self.state = 'up'
  else
    self.state = (auto_start == false) ? 'unmonitored' : 'down' # we need to check for false value
  end
end

#dispatch!(event, reason = nil) ⇒ Object

State machine methods



175
176
177
178
179
180
 
# File 'lib/bluepill/process.rb', line 175

def dispatch!(event, reason = nil)
  @event_mutex.synchronize do
    @statistics.record_event(event, reason)
    send("#{event}")
  end
end

#handle_user_command(cmd) ⇒ Object 



247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
 
# File 'lib/bluepill/process.rb', line 247

def handle_user_command(cmd)
  case cmd
  when 'start'
    if self.process_running?(true)
      logger.warning('Refusing to re-run start command on an already running process.')
    else
      dispatch!(:start, 'user initiated')
    end
  when 'stop'
    stop_process
    dispatch!(:unmonitor, 'user initiated')
  when 'restart'
    restart_process
  when 'unmonitor'
    # When the user issues an unmonitor cmd, reset any triggers so that
    # scheduled events gets cleared
    triggers.each(&:reset!)
    dispatch!(:unmonitor, 'user initiated')
  end
end

#monitor_children? ⇒ Boolean

Returns:

  • (Boolean) 


406
407
408
 
# File 'lib/bluepill/process.rb', line 406

def monitor_children?
  !!monitor_children
end

#notify_triggers(transition) ⇒ Object 



197
198
199
200
201
202
203
204
205
206
 
# File 'lib/bluepill/process.rb', line 197

def notify_triggers(transition)
  triggers.each do |trigger|
    begin
      trigger.notify(transition)
    rescue => e
      logger.err e.backtrace
      raise e
    end
  end
end

#pid_from_command ⇒ Object 



442
443
444
445
 
# File 'lib/bluepill/process.rb', line 442

def pid_from_command
  pid = `#{pid_command}`.strip
  (pid =~ /\A\d+\z/) ? pid.to_i : nil
end

#pid_from_file ⇒ Object 



427
428
429
430
431
432
433
434
435
436
437
438
439
440
 
# File 'lib/bluepill/process.rb', line 427

def pid_from_file
  return @actual_pid if cache_actual_pid? && @actual_pid
  @actual_pid = begin
    if pid_file
      if File.exist?(pid_file)
        str = File.read(pid_file)
        str.to_i if str.size > 0
      else
        logger.warning("pid_file #{pid_file} does not exist or cannot be read")
        nil
      end
    end
  end
end

#pre_start_process ⇒ Object 



306
307
308
309
310
311
312
313
 
# File 'lib/bluepill/process.rb', line 306

def pre_start_process
  return unless pre_start_command
  logger.warning "Executing pre start command: #{pre_start_command}"
  result = System.execute_blocking(pre_start_command, system_command_options)
  return if result[:exit_code].zero?
  logger.warning 'Pre start command execution returned non-zero exit code:'
  logger.warning result.inspect
end

#prepare_command(command) ⇒ Object 



493
494
495
 
# File 'lib/bluepill/process.rb', line 493

def prepare_command(command)
  command.to_s.gsub('{{PID}}', actual_pid.to_s)
end

#process_running?(force = false) ⇒ Boolean

System Process Methods

Returns:

  • (Boolean) 


269
270
271
272
273
274
275
276
 
# File 'lib/bluepill/process.rb', line 269

def process_running?(force = false)
  @process_running = nil if force # clear existing state if forced

  @process_running ||= signal_process(0)
  # the process isn't running, so we should clear the PID
  clear_pid unless @process_running
  @process_running
end

#record_transition(transition) ⇒ Object 



182
183
184
185
186
187
188
189
190
191
192
193
194
195
 
# File 'lib/bluepill/process.rb', line 182

def record_transition(transition)
  return if transition.loopback?
  @transitioned = true

  # When a process changes state, we should clear the memory of all the watches
  watches.each(&:clear_history!)

  # Also, when a process changes state, we should re-populate its child list
  if self.monitor_children?
    logger.warning 'Clearing child list'
    children.clear
  end
  logger.info "Going from #{transition.from_name} => #{transition.to_name}"
end

#refresh_children! ⇒ Object 



471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
 
# File 'lib/bluepill/process.rb', line 471

def refresh_children!
  # First prune the list of dead children
  @children.delete_if { |child| !child.process_running?(true) }

  # Add new found children to the list
  new_children_pids = System.get_children(actual_pid) - @children.collect(&:actual_pid)

  unless new_children_pids.empty?
    logger.info "Existing children: #{@children.collect(&:actual_pid).join(',')}. Got new children: #{new_children_pids.inspect} for #{actual_pid}"
  end

  # Construct a new process wrapper for each new found children
  new_children_pids.each do |child_pid|
    ProcessJournal.append_pid_to_journal(name, child_pid) if daemonize?
    child_name = "<child(pid:#{child_pid})>"
    logger = self.logger.prefix_with(child_name)

    child = child_process_factory.create_child_process(child_name, child_pid, logger)
    @children << child
  end
end

#restart_process ⇒ Object 



369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
 
# File 'lib/bluepill/process.rb', line 369

def restart_process
  if restart_command
    cmd = prepare_command(restart_command)

    logger.warning "Executing restart command: #{cmd}"

    with_timeout(restart_grace_time, 'restart') do
      result = System.execute_blocking(cmd, system_command_options)

      unless result[:exit_code].zero?
        logger.warning 'Restart command execution returned non-zero exit code:'
        logger.warning result.inspect
      end
    end

    skip_ticks_for(restart_grace_time)
  else
    logger.warning 'No restart_command specified. Must stop and start to restart'
    stop_process
    start_process
  end
end

#run_watches ⇒ Object 



217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
 
# File 'lib/bluepill/process.rb', line 217

def run_watches
  now = Time.now.to_i

  threads = watches.collect do |watch|
    [watch, Thread.new { Thread.current[:events] = watch.run(actual_pid, now) }]
  end

  @transitioned = false

  threads.each_with_object([]) do |(watch, thread), events|
    thread.join
    next if thread[:events].size.zero?
    logger.info "#{watch.name} dispatched: #{thread[:events].join(',')}"
    thread[:events].each do |event|
      events << [event, watch.to_s]
    end
  end.each do |event, reason| # rubocop:disable Style/MultilineBlockChain
    break if @transitioned
    self.dispatch!(event, reason)
  end
end

#signal_process(code) ⇒ Object 



410
411
412
413
414
415
416
417
 
# File 'lib/bluepill/process.rb', line 410

def signal_process(code)
  code = code.to_s.upcase if code.is_a?(String) || code.is_a?(Symbol)
  ::Process.kill(code, actual_pid)
  true
rescue => e
  logger.err "Failed to signal process #{actual_pid} with code #{code}: #{e}"
  false
end

#skip_ticks_for(seconds) ⇒ Object

Internal State Methods



461
462
463
464
465
 
# File 'lib/bluepill/process.rb', line 461

def skip_ticks_for(seconds)
  # TODO: should this be addative or longest wins?
  #       i.e. if two calls for skip_ticks_for come in for 5 and 10, should it skip for 10 or 15?
  self.skip_ticks_until = (skip_ticks_until || Time.now.to_i) + seconds.to_i
end

#skipping_ticks? ⇒ Boolean

Returns:

  • (Boolean) 


467
468
469
 
# File 'lib/bluepill/process.rb', line 467

def skipping_ticks?
  skip_ticks_until && skip_ticks_until > Time.now.to_i
end

#start_process ⇒ Object 



278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
 
# File 'lib/bluepill/process.rb', line 278

def start_process
  ProcessJournal.kill_all_from_journal(name) # be sure nothing else is running from previous runs
  pre_start_process
  logger.warning "Executing start command: #{start_command}"
  if self.daemonize?
    daemon_id = System.daemonize(start_command, system_command_options)
    if daemon_id
      ProcessJournal.append_pid_to_journal(name, daemon_id)
      children.each do|child|
        ProcessJournal.append_pid_to_journal(name, child.actual_pid)
      end if self.monitor_children?
    end
    daemon_id
  else
    # This is a self-daemonizing process
    with_timeout(start_grace_time, on_start_timeout) do
      result = System.execute_blocking(start_command, system_command_options)

      unless result[:exit_code].zero?
        logger.warning 'Start command execution returned non-zero exit code:'
        logger.warning result.inspect
      end
    end
  end

  skip_ticks_for(start_grace_time)
end

#stop_process ⇒ Object 



315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
 
# File 'lib/bluepill/process.rb', line 315

def stop_process
  if monitor_children
    System.get_children(actual_pid).each do |child_pid|
      ProcessJournal.append_pid_to_journal(name, child_pid)
    end
  end

  if stop_command
    cmd = prepare_command(stop_command)
    logger.warning "Executing stop command: #{cmd}"

    with_timeout(stop_grace_time, 'stop') do
      result = System.execute_blocking(cmd, system_command_options)

      unless result[:exit_code].zero?
        logger.warning 'Stop command execution returned non-zero exit code:'
        logger.warning result.inspect
      end
    end
    cleanup_process
  elsif stop_signals
    # issue stop signals with configurable delay between each
    logger.warning "Sending stop signals to #{actual_pid}"
    @threads << Thread.new(self, stop_signals.clone) do |process, stop_signals|
      signal = stop_signals.shift
      logger.info "Sending signal #{signal} to #{process.actual_pid}"
      process.signal_process(signal) # send first signal

      until stop_signals.empty?
        # we already checked to make sure stop_signals had an odd number of items
        delay = stop_signals.shift
        signal = stop_signals.shift

        logger.debug "Sleeping for #{delay} seconds"
        sleep delay
        # break unless signal_process(0) #break unless the process can be reached
        unless process.signal_process(0)
          logger.debug 'Process has terminated.'
          break
        end
        logger.info "Sending signal #{signal} to #{process.actual_pid}"
        process.signal_process(signal)
      end
      cleanup_process
    end
  else
    logger.warning "Executing default stop command. Sending TERM signal to #{actual_pid}"
    signal_process('TERM')
    cleanup_process
  end

  skip_ticks_for(stop_grace_time)
end

#system_command_options ⇒ Object 



497
498
499
500
501
502
503
504
505
506
507
508
509
510
 
# File 'lib/bluepill/process.rb', line 497

def system_command_options
  {
    uid: uid,
    gid: gid,
    working_dir: working_dir,
    environment: environment,
    pid_file: pid_file,
    logger: logger,
    stdin: stdin,
    stdout: stdout,
    stderr: stderr,
    supplementary_groups: supplementary_groups,
  }
end

#tick ⇒ Object 



147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
 
# File 'lib/bluepill/process.rb', line 147

def tick
  return if self.skipping_ticks?
  self.skip_ticks_until = nil

  # clear the memoization per tick
  @process_running = nil

  # Deal with thread cleanup here since the stopping state isn't used
  clean_threads if self.unmonitored?

  # run state machine transitions
  super

  return unless self.up?
  run_watches

  return unless self.monitor_children?
  refresh_children!
  children.each(&:tick)
end

#unlink_pid ⇒ Object 



456
457
458
 
# File 'lib/bluepill/process.rb', line 456

def unlink_pid
  System.delete_if_exists(pid_file)
end

#with_timeout(secs, next_state = nil, &blk) ⇒ Object 



512
513
514
515
516
517
518
519
520
 
# File 'lib/bluepill/process.rb', line 512

def with_timeout(secs, next_state = nil, &blk)
  # Attempt to execute the passed block. If the block takes
  # too long, transition to the indicated next state.
  Timeout.timeout(secs.to_f, &blk)
rescue Timeout::Error
  logger.err 'Execution is taking longer than expected.'
  logger.err 'Did you forget to tell bluepill to daemonize this process?'
  dispatch!(next_state)
end