efficiency20-delayed_job 1.8.51
Sign up to get free protection for your applications and to get access to all the features.
- data/.gitignore +1 -0
- data/MIT-LICENSE +20 -0
- data/README.textile +184 -0
- data/Rakefile +34 -0
- data/VERSION +1 -0
- data/contrib/delayed_job.monitrc +14 -0
- data/efficiency20-delayed_job.gemspec +70 -0
- data/generators/delayed_job/delayed_job_generator.rb +22 -0
- data/generators/delayed_job/templates/migration.rb +20 -0
- data/generators/delayed_job/templates/script +5 -0
- data/init.rb +1 -0
- data/lib/delayed/command.rb +103 -0
- data/lib/delayed/job.rb +296 -0
- data/lib/delayed/message_sending.rb +22 -0
- data/lib/delayed/performable_method.rb +55 -0
- data/lib/delayed/recipes.rb +31 -0
- data/lib/delayed/tasks.rb +15 -0
- data/lib/delayed/worker.rb +77 -0
- data/lib/delayed_job.rb +13 -0
- data/recipes/delayed_job.rb +1 -0
- data/spec/database.rb +43 -0
- data/spec/delayed_method_spec.rb +150 -0
- data/spec/job_spec.rb +419 -0
- data/spec/story_spec.rb +17 -0
- data/tasks/jobs.rake +1 -0
- metadata +105 -0
data/lib/delayed/job.rb
ADDED
@@ -0,0 +1,296 @@
|
|
1
|
+
require 'timeout'
|
2
|
+
|
3
|
+
module Delayed
|
4
|
+
|
5
|
+
class DeserializationError < StandardError
|
6
|
+
end
|
7
|
+
|
8
|
+
# A job object that is persisted to the database.
|
9
|
+
# Contains the work object as a YAML field.
|
10
|
+
class Job < ActiveRecord::Base
|
11
|
+
@@max_attempts = 25
|
12
|
+
@@max_run_time = 4.hours
|
13
|
+
|
14
|
+
cattr_accessor :max_attempts, :max_run_time
|
15
|
+
|
16
|
+
set_table_name :delayed_jobs
|
17
|
+
|
18
|
+
# By default failed jobs are destroyed after too many attempts.
|
19
|
+
# If you want to keep them around (perhaps to inspect the reason
|
20
|
+
# for the failure), set this to false.
|
21
|
+
cattr_accessor :destroy_failed_jobs
|
22
|
+
self.destroy_failed_jobs = true
|
23
|
+
|
24
|
+
# Every worker has a unique name which by default is the pid of the process.
|
25
|
+
# There are some advantages to overriding this with something which survives worker retarts:
|
26
|
+
# Workers can safely resume working on tasks which are locked by themselves. The worker will assume that it crashed before.
|
27
|
+
@@worker_name = nil
|
28
|
+
|
29
|
+
def self.worker_name
|
30
|
+
return @@worker_name unless @@worker_name.nil?
|
31
|
+
"host:#{Socket.gethostname} pid:#{Process.pid}" rescue "pid:#{Process.pid}"
|
32
|
+
end
|
33
|
+
|
34
|
+
def self.worker_name=(val)
|
35
|
+
@@worker_name = val
|
36
|
+
end
|
37
|
+
|
38
|
+
def worker_name
|
39
|
+
self.class.worker_name
|
40
|
+
end
|
41
|
+
|
42
|
+
def worker_name=(val)
|
43
|
+
@@worker_name = val
|
44
|
+
end
|
45
|
+
|
46
|
+
NextTaskSQL = '(run_at <= ? AND (locked_at IS NULL OR locked_at < ?) OR (locked_by = ?)) AND failed_at IS NULL'
|
47
|
+
NextTaskOrder = 'priority DESC, run_at ASC'
|
48
|
+
|
49
|
+
ParseObjectFromYaml = /\!ruby\/\w+\:([^\s]+)/
|
50
|
+
|
51
|
+
cattr_accessor :min_priority, :max_priority
|
52
|
+
self.min_priority = nil
|
53
|
+
self.max_priority = nil
|
54
|
+
|
55
|
+
# When a worker is exiting, make sure we don't have any locked jobs.
|
56
|
+
def self.clear_locks!
|
57
|
+
update_all("locked_by = null, locked_at = null", ["locked_by = ?", worker_name])
|
58
|
+
end
|
59
|
+
|
60
|
+
def failed?
|
61
|
+
failed_at
|
62
|
+
end
|
63
|
+
alias_method :failed, :failed?
|
64
|
+
|
65
|
+
def payload_object
|
66
|
+
@payload_object ||= deserialize(self['handler'])
|
67
|
+
end
|
68
|
+
|
69
|
+
def name
|
70
|
+
@name ||= begin
|
71
|
+
payload = payload_object
|
72
|
+
if payload.respond_to?(:display_name)
|
73
|
+
payload.display_name
|
74
|
+
else
|
75
|
+
payload.class.name
|
76
|
+
end
|
77
|
+
end
|
78
|
+
end
|
79
|
+
|
80
|
+
def payload_object=(object)
|
81
|
+
self['handler'] = object.to_yaml
|
82
|
+
end
|
83
|
+
|
84
|
+
# Reschedule the job in the future (when a job fails).
|
85
|
+
# Uses an exponential scale depending on the number of failed attempts.
|
86
|
+
def reschedule(message, backtrace = [], time = nil)
|
87
|
+
self.last_error = message + "\n" + backtrace.join("\n")
|
88
|
+
|
89
|
+
if (self.attempts += 1) < max_attempts
|
90
|
+
time ||= Job.db_time_now + (attempts ** 4) + 5
|
91
|
+
|
92
|
+
self.run_at = time
|
93
|
+
self.unlock
|
94
|
+
save!
|
95
|
+
else
|
96
|
+
logger.info "* [JOB] PERMANENTLY removing #{self.name} because of #{attempts} consecutive failures."
|
97
|
+
destroy_failed_jobs ? destroy : update_attribute(:failed_at, Delayed::Job.db_time_now)
|
98
|
+
end
|
99
|
+
end
|
100
|
+
|
101
|
+
|
102
|
+
# Try to run one job. Returns true/false (work done/work failed) or nil if job can't be locked.
|
103
|
+
def run_with_lock(max_run_time, worker_name)
|
104
|
+
logger.info "* [JOB] acquiring lock on #{name}"
|
105
|
+
unless lock_exclusively!(max_run_time, worker_name)
|
106
|
+
# We did not get the lock, some other worker process must have
|
107
|
+
logger.warn "* [JOB] failed to acquire exclusive lock for #{name}"
|
108
|
+
return nil # no work done
|
109
|
+
end
|
110
|
+
|
111
|
+
begin
|
112
|
+
runtime = Benchmark.realtime do
|
113
|
+
Timeout.timeout(max_run_time.to_i) { invoke_job }
|
114
|
+
destroy
|
115
|
+
end
|
116
|
+
# TODO: warn if runtime > max_run_time ?
|
117
|
+
logger.info "* [JOB] #{name} completed after %.4f" % runtime
|
118
|
+
return true # did work
|
119
|
+
rescue Exception => e
|
120
|
+
reschedule e.message, e.backtrace
|
121
|
+
log_exception(e)
|
122
|
+
return false # work failed
|
123
|
+
end
|
124
|
+
end
|
125
|
+
|
126
|
+
# Add a job to the queue
|
127
|
+
def self.enqueue(*args, &block)
|
128
|
+
object = block_given? ? EvaledJob.new(&block) : args.shift
|
129
|
+
|
130
|
+
unless object.respond_to?(:perform) || block_given?
|
131
|
+
raise ArgumentError, 'Cannot enqueue items which do not respond to perform'
|
132
|
+
end
|
133
|
+
|
134
|
+
priority = args.first || 0
|
135
|
+
run_at = args[1]
|
136
|
+
|
137
|
+
Job.create(:payload_object => object, :priority => priority.to_i, :run_at => run_at)
|
138
|
+
end
|
139
|
+
|
140
|
+
# Find a few candidate jobs to run (in case some immediately get locked by others).
|
141
|
+
def self.find_available(limit = 5, max_run_time = max_run_time)
|
142
|
+
|
143
|
+
time_now = db_time_now
|
144
|
+
|
145
|
+
sql = NextTaskSQL.dup
|
146
|
+
|
147
|
+
conditions = [time_now, time_now - max_run_time, worker_name]
|
148
|
+
|
149
|
+
if self.min_priority
|
150
|
+
sql << ' AND (priority >= ?)'
|
151
|
+
conditions << min_priority
|
152
|
+
end
|
153
|
+
|
154
|
+
if self.max_priority
|
155
|
+
sql << ' AND (priority <= ?)'
|
156
|
+
conditions << max_priority
|
157
|
+
end
|
158
|
+
|
159
|
+
conditions.unshift(sql)
|
160
|
+
|
161
|
+
ActiveRecord::Base.silence do
|
162
|
+
find(:all, :conditions => conditions, :order => NextTaskOrder, :limit => limit)
|
163
|
+
end
|
164
|
+
end
|
165
|
+
|
166
|
+
# Run the next job we can get an exclusive lock on.
|
167
|
+
# If no jobs are left we return nil
|
168
|
+
def self.reserve_and_run_one_job(max_run_time = max_run_time)
|
169
|
+
|
170
|
+
# We get up to 5 jobs from the db. In case we cannot get exclusive access to a job we try the next.
|
171
|
+
# this leads to a more even distribution of jobs across the worker processes
|
172
|
+
find_available(5, max_run_time).each do |job|
|
173
|
+
t = job.run_with_lock(max_run_time, worker_name)
|
174
|
+
return t unless t == nil # return if we did work (good or bad)
|
175
|
+
end
|
176
|
+
|
177
|
+
nil # we didn't do any work, all 5 were not lockable
|
178
|
+
end
|
179
|
+
|
180
|
+
# Lock this job for this worker.
|
181
|
+
# Returns true if we have the lock, false otherwise.
|
182
|
+
def lock_exclusively!(max_run_time, worker = worker_name)
|
183
|
+
now = self.class.db_time_now
|
184
|
+
affected_rows = if locked_by != worker
|
185
|
+
# We don't own this job so we will update the locked_by name and the locked_at
|
186
|
+
self.class.update_all(["locked_at = ?, locked_by = ?", now, worker], ["id = ? and (locked_at is null or locked_at < ?) and (run_at <= ?)", id, (now - max_run_time.to_i), now])
|
187
|
+
else
|
188
|
+
# We already own this job, this may happen if the job queue crashes.
|
189
|
+
# Simply resume and update the locked_at
|
190
|
+
self.class.update_all(["locked_at = ?", now], ["id = ? and locked_by = ?", id, worker])
|
191
|
+
end
|
192
|
+
if affected_rows == 1
|
193
|
+
self.locked_at = now
|
194
|
+
self.locked_by = worker
|
195
|
+
return true
|
196
|
+
else
|
197
|
+
return false
|
198
|
+
end
|
199
|
+
end
|
200
|
+
|
201
|
+
# Unlock this job (note: not saved to DB)
|
202
|
+
def unlock
|
203
|
+
self.locked_at = nil
|
204
|
+
self.locked_by = nil
|
205
|
+
end
|
206
|
+
|
207
|
+
# This is a good hook if you need to report job processing errors in additional or different ways
|
208
|
+
def log_exception(error)
|
209
|
+
logger.error "* [JOB] #{name} failed with #{error.class.name}: #{error.message} - #{attempts} failed attempts"
|
210
|
+
logger.error(error)
|
211
|
+
end
|
212
|
+
|
213
|
+
# Do num jobs and return stats on success/failure.
|
214
|
+
# Exit early if interrupted.
|
215
|
+
def self.work_off(num = 100)
|
216
|
+
success, failure = 0, 0
|
217
|
+
|
218
|
+
num.times do
|
219
|
+
case self.reserve_and_run_one_job
|
220
|
+
when true
|
221
|
+
success += 1
|
222
|
+
when false
|
223
|
+
failure += 1
|
224
|
+
else
|
225
|
+
break # leave if no work could be done
|
226
|
+
end
|
227
|
+
break if $exit # leave if we're exiting
|
228
|
+
end
|
229
|
+
|
230
|
+
return [success, failure]
|
231
|
+
end
|
232
|
+
|
233
|
+
# Moved into its own method so that new_relic can trace it.
|
234
|
+
def invoke_job
|
235
|
+
payload_object.perform
|
236
|
+
end
|
237
|
+
|
238
|
+
private
|
239
|
+
|
240
|
+
def deserialize(source)
|
241
|
+
handler = YAML.load(source) rescue nil
|
242
|
+
|
243
|
+
unless handler.respond_to?(:perform)
|
244
|
+
if handler.nil? && source =~ ParseObjectFromYaml
|
245
|
+
handler_class = $1
|
246
|
+
end
|
247
|
+
attempt_to_load(handler_class || handler.class)
|
248
|
+
handler = YAML.load(source)
|
249
|
+
end
|
250
|
+
|
251
|
+
return handler if handler.respond_to?(:perform)
|
252
|
+
|
253
|
+
raise DeserializationError,
|
254
|
+
'Job failed to load: Unknown handler. Try to manually require the appropriate file.'
|
255
|
+
rescue TypeError, LoadError, NameError => e
|
256
|
+
raise DeserializationError,
|
257
|
+
"Job failed to load: #{e.message}. Try to manually require the required file."
|
258
|
+
end
|
259
|
+
|
260
|
+
# Constantize the object so that ActiveSupport can attempt
|
261
|
+
# its auto loading magic. Will raise LoadError if not successful.
|
262
|
+
def attempt_to_load(klass)
|
263
|
+
klass.constantize
|
264
|
+
end
|
265
|
+
|
266
|
+
# Get the current time (GMT or local depending on DB)
|
267
|
+
# Note: This does not ping the DB to get the time, so all your clients
|
268
|
+
# must have syncronized clocks.
|
269
|
+
def self.db_time_now
|
270
|
+
if (defined? Time.zone) && Time.zone
|
271
|
+
Time.zone.now
|
272
|
+
elsif ActiveRecord::Base.default_timezone == :utc
|
273
|
+
Time.now.utc
|
274
|
+
else
|
275
|
+
Time.now
|
276
|
+
end
|
277
|
+
end
|
278
|
+
|
279
|
+
protected
|
280
|
+
|
281
|
+
def before_save
|
282
|
+
self.run_at ||= self.class.db_time_now
|
283
|
+
end
|
284
|
+
|
285
|
+
end
|
286
|
+
|
287
|
+
class EvaledJob
|
288
|
+
def initialize
|
289
|
+
@job = yield
|
290
|
+
end
|
291
|
+
|
292
|
+
def perform
|
293
|
+
eval(@job)
|
294
|
+
end
|
295
|
+
end
|
296
|
+
end
|
@@ -0,0 +1,22 @@
|
|
1
|
+
module Delayed
|
2
|
+
module MessageSending
|
3
|
+
def send_later(method, *args)
|
4
|
+
Delayed::Job.enqueue Delayed::PerformableMethod.new(self, method.to_sym, args)
|
5
|
+
end
|
6
|
+
|
7
|
+
def send_at(time, method, *args)
|
8
|
+
Delayed::Job.enqueue(Delayed::PerformableMethod.new(self, method.to_sym, args), 0, time)
|
9
|
+
end
|
10
|
+
|
11
|
+
module ClassMethods
|
12
|
+
def handle_asynchronously(method)
|
13
|
+
aliased_method, punctuation = method.to_s.sub(/([?!=])$/, ''), $1
|
14
|
+
with_method, without_method = "#{aliased_method}_with_send_later#{punctuation}", "#{aliased_method}_without_send_later#{punctuation}"
|
15
|
+
define_method(with_method) do |*args|
|
16
|
+
send_later(without_method, *args)
|
17
|
+
end
|
18
|
+
alias_method_chain method, :send_later
|
19
|
+
end
|
20
|
+
end
|
21
|
+
end
|
22
|
+
end
|
@@ -0,0 +1,55 @@
|
|
1
|
+
module Delayed
|
2
|
+
class PerformableMethod < Struct.new(:object, :method, :args)
|
3
|
+
CLASS_STRING_FORMAT = /^CLASS\:([A-Z][\w\:]+)$/
|
4
|
+
AR_STRING_FORMAT = /^AR\:([A-Z][\w\:]+)\:(\d+)$/
|
5
|
+
|
6
|
+
def initialize(object, method, args)
|
7
|
+
raise NoMethodError, "undefined method `#{method}' for #{object.inspect}" unless object.respond_to?(method)
|
8
|
+
|
9
|
+
self.object = dump(object)
|
10
|
+
self.args = args.map { |a| dump(a) }
|
11
|
+
self.method = method.to_sym
|
12
|
+
end
|
13
|
+
|
14
|
+
def display_name
|
15
|
+
case self.object
|
16
|
+
when CLASS_STRING_FORMAT then "#{$1}.#{method}"
|
17
|
+
when AR_STRING_FORMAT then "#{$1}##{method}"
|
18
|
+
else "Unknown##{method}"
|
19
|
+
end
|
20
|
+
end
|
21
|
+
|
22
|
+
def perform
|
23
|
+
load(object).send(method, *args.map{|a| load(a)})
|
24
|
+
rescue ActiveRecord::RecordNotFound
|
25
|
+
# We cannot do anything about objects which were deleted in the meantime
|
26
|
+
true
|
27
|
+
end
|
28
|
+
|
29
|
+
private
|
30
|
+
|
31
|
+
def load(arg)
|
32
|
+
case arg
|
33
|
+
when CLASS_STRING_FORMAT then $1.constantize
|
34
|
+
when AR_STRING_FORMAT then $1.constantize.find($2)
|
35
|
+
else arg
|
36
|
+
end
|
37
|
+
end
|
38
|
+
|
39
|
+
def dump(arg)
|
40
|
+
case arg
|
41
|
+
when Class then class_to_string(arg)
|
42
|
+
when ActiveRecord::Base then ar_to_string(arg)
|
43
|
+
else arg
|
44
|
+
end
|
45
|
+
end
|
46
|
+
|
47
|
+
def ar_to_string(obj)
|
48
|
+
"AR:#{obj.class}:#{obj.id}"
|
49
|
+
end
|
50
|
+
|
51
|
+
def class_to_string(obj)
|
52
|
+
"CLASS:#{obj.name}"
|
53
|
+
end
|
54
|
+
end
|
55
|
+
end
|
@@ -0,0 +1,31 @@
|
|
1
|
+
# Capistrano Recipes for managing delayed_job
|
2
|
+
#
|
3
|
+
# Add these callbacks to have the delayed_job process restart when the server
|
4
|
+
# is restarted:
|
5
|
+
#
|
6
|
+
# after "deploy:stop", "delayed_job:stop"
|
7
|
+
# after "deploy:start", "delayed_job:start"
|
8
|
+
# after "deploy:restart", "delayed_job:restart"
|
9
|
+
|
10
|
+
Capistrano::Configuration.instance.load do
|
11
|
+
namespace :delayed_job do
|
12
|
+
def rails_env
|
13
|
+
fetch(:rails_env, false) ? "RAILS_ENV=#{fetch(:rails_env)}" : ''
|
14
|
+
end
|
15
|
+
|
16
|
+
desc "Stop the delayed_job process"
|
17
|
+
task :stop, :roles => :app do
|
18
|
+
run "cd #{current_path};#{rails_env} script/delayed_job stop"
|
19
|
+
end
|
20
|
+
|
21
|
+
desc "Start the delayed_job process"
|
22
|
+
task :start, :roles => :app do
|
23
|
+
run "cd #{current_path};#{rails_env} script/delayed_job start"
|
24
|
+
end
|
25
|
+
|
26
|
+
desc "Restart the delayed_job process"
|
27
|
+
task :restart, :roles => :app do
|
28
|
+
run "cd #{current_path};#{rails_env} script/delayed_job restart"
|
29
|
+
end
|
30
|
+
end
|
31
|
+
end
|
@@ -0,0 +1,15 @@
|
|
1
|
+
# Re-definitions are appended to existing tasks
|
2
|
+
task :environment
|
3
|
+
task :merb_env
|
4
|
+
|
5
|
+
namespace :jobs do
|
6
|
+
desc "Clear the delayed_job queue."
|
7
|
+
task :clear => [:merb_env, :environment] do
|
8
|
+
Delayed::Job.delete_all
|
9
|
+
end
|
10
|
+
|
11
|
+
desc "Start a delayed_job worker."
|
12
|
+
task :work => [:merb_env, :environment] do
|
13
|
+
Delayed::Worker.new(:min_priority => ENV['MIN_PRIORITY'], :max_priority => ENV['MAX_PRIORITY']).start
|
14
|
+
end
|
15
|
+
end
|
@@ -0,0 +1,77 @@
|
|
1
|
+
module Delayed
|
2
|
+
class Worker
|
3
|
+
@@sleep_delay = 5
|
4
|
+
|
5
|
+
cattr_accessor :sleep_delay
|
6
|
+
|
7
|
+
cattr_accessor :logger
|
8
|
+
self.logger = if defined?(Merb::Logger)
|
9
|
+
Merb.logger
|
10
|
+
elsif defined?(RAILS_DEFAULT_LOGGER)
|
11
|
+
RAILS_DEFAULT_LOGGER
|
12
|
+
end
|
13
|
+
|
14
|
+
# name_prefix is ignored if name is set directly
|
15
|
+
attr_accessor :name_prefix
|
16
|
+
|
17
|
+
def job_max_run_time
|
18
|
+
Delayed::Job.max_run_time
|
19
|
+
end
|
20
|
+
|
21
|
+
# Every worker has a unique name which by default is the pid of the process.
|
22
|
+
# There are some advantages to overriding this with something which survives worker retarts:
|
23
|
+
# Workers can safely resume working on tasks which are locked by themselves. The worker will assume that it crashed before.
|
24
|
+
def name
|
25
|
+
return @name unless @name.nil?
|
26
|
+
"#{@name_prefix}host:#{Socket.gethostname} pid:#{Process.pid}" rescue "#{@name_prefix}pid:#{Process.pid}"
|
27
|
+
end
|
28
|
+
|
29
|
+
# Sets the name of the worker.
|
30
|
+
# Setting the name to nil will reset the default worker name
|
31
|
+
def name=(val)
|
32
|
+
@name = val
|
33
|
+
end
|
34
|
+
|
35
|
+
def initialize(options={})
|
36
|
+
@quiet = options[:quiet]
|
37
|
+
Delayed::Job.min_priority = options[:min_priority] if options.has_key?(:min_priority)
|
38
|
+
Delayed::Job.max_priority = options[:max_priority] if options.has_key?(:max_priority)
|
39
|
+
end
|
40
|
+
|
41
|
+
def start
|
42
|
+
say "*** Starting job worker #{Delayed::Job.worker_name}"
|
43
|
+
|
44
|
+
trap('TERM') { say 'Exiting...'; $exit = true }
|
45
|
+
trap('INT') { say 'Exiting...'; $exit = true }
|
46
|
+
|
47
|
+
loop do
|
48
|
+
result = nil
|
49
|
+
|
50
|
+
realtime = Benchmark.realtime do
|
51
|
+
result = Delayed::Job.work_off
|
52
|
+
end
|
53
|
+
|
54
|
+
count = result.sum
|
55
|
+
|
56
|
+
break if $exit
|
57
|
+
|
58
|
+
if count.zero?
|
59
|
+
sleep(@@sleep_delay)
|
60
|
+
else
|
61
|
+
say "#{count} jobs processed at %.4f j/s, %d failed ..." % [count / realtime, result.last]
|
62
|
+
end
|
63
|
+
|
64
|
+
break if $exit
|
65
|
+
end
|
66
|
+
|
67
|
+
ensure
|
68
|
+
Delayed::Job.clear_locks!
|
69
|
+
end
|
70
|
+
|
71
|
+
def say(text)
|
72
|
+
puts text unless @quiet
|
73
|
+
logger.info text if logger
|
74
|
+
end
|
75
|
+
|
76
|
+
end
|
77
|
+
end
|