inst-jobs 2.3.2 → 2.4.2
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/db/migrate/20101216224513_create_delayed_jobs.rb +9 -7
- data/db/migrate/20110531144916_cleanup_delayed_jobs_indexes.rb +8 -13
- data/db/migrate/20110610213249_optimize_delayed_jobs.rb +8 -8
- data/db/migrate/20110831210257_add_delayed_jobs_next_in_strand.rb +25 -25
- data/db/migrate/20120510004759_delayed_jobs_delete_trigger_lock_for_update.rb +4 -8
- data/db/migrate/20120531150712_drop_psql_jobs_pop_fn.rb +1 -3
- data/db/migrate/20120607164022_delayed_jobs_use_advisory_locks.rb +11 -15
- data/db/migrate/20120607181141_index_jobs_on_locked_by.rb +1 -1
- data/db/migrate/20120608191051_add_jobs_run_at_index.rb +2 -2
- data/db/migrate/20120927184213_change_delayed_jobs_handler_to_text.rb +1 -1
- data/db/migrate/20140505215510_copy_failed_jobs_original_id.rb +2 -3
- data/db/migrate/20150807133223_add_max_concurrent_to_jobs.rb +9 -13
- data/db/migrate/20151210162949_improve_max_concurrent.rb +4 -8
- data/db/migrate/20161206323555_add_back_default_string_limits_jobs.rb +3 -2
- data/db/migrate/20181217155351_speed_up_max_concurrent_triggers.rb +13 -17
- data/db/migrate/20200330230722_add_id_to_get_delayed_jobs_index.rb +8 -8
- data/db/migrate/20200824222232_speed_up_max_concurrent_delete_trigger.rb +72 -77
- data/db/migrate/20200825011002_add_strand_order_override.rb +93 -97
- data/db/migrate/20210809145804_add_n_strand_index.rb +12 -0
- data/db/migrate/20210812210128_add_singleton_column.rb +203 -0
- data/exe/inst_jobs +3 -2
- data/lib/delayed/backend/active_record.rb +182 -148
- data/lib/delayed/backend/base.rb +79 -74
- data/lib/delayed/batch.rb +11 -9
- data/lib/delayed/cli.rb +98 -84
- data/lib/delayed/core_ext/kernel.rb +4 -2
- data/lib/delayed/daemon.rb +70 -74
- data/lib/delayed/job_tracking.rb +26 -25
- data/lib/delayed/lifecycle.rb +27 -24
- data/lib/delayed/log_tailer.rb +17 -17
- data/lib/delayed/logging.rb +13 -16
- data/lib/delayed/message_sending.rb +42 -51
- data/lib/delayed/performable_method.rb +5 -7
- data/lib/delayed/periodic.rb +66 -65
- data/lib/delayed/plugin.rb +2 -4
- data/lib/delayed/pool.rb +198 -193
- data/lib/delayed/server/helpers.rb +6 -6
- data/lib/delayed/server.rb +51 -54
- data/lib/delayed/settings.rb +93 -81
- data/lib/delayed/testing.rb +21 -22
- data/lib/delayed/version.rb +1 -1
- data/lib/delayed/work_queue/in_process.rb +21 -18
- data/lib/delayed/work_queue/parent_process/client.rb +54 -55
- data/lib/delayed/work_queue/parent_process/server.rb +215 -209
- data/lib/delayed/work_queue/parent_process.rb +52 -53
- data/lib/delayed/worker/consul_health_check.rb +21 -19
- data/lib/delayed/worker/health_check.rb +21 -12
- data/lib/delayed/worker/null_health_check.rb +3 -1
- data/lib/delayed/worker/process_helper.rb +8 -9
- data/lib/delayed/worker.rb +271 -265
- data/lib/delayed/yaml_extensions.rb +12 -10
- data/lib/delayed_job.rb +37 -38
- data/lib/inst-jobs.rb +1 -1
- data/spec/active_record_job_spec.rb +128 -135
- data/spec/delayed/cli_spec.rb +7 -7
- data/spec/delayed/daemon_spec.rb +8 -8
- data/spec/delayed/message_sending_spec.rb +16 -9
- data/spec/delayed/periodic_spec.rb +13 -12
- data/spec/delayed/server_spec.rb +38 -38
- data/spec/delayed/settings_spec.rb +26 -25
- data/spec/delayed/work_queue/in_process_spec.rb +7 -7
- data/spec/delayed/work_queue/parent_process/client_spec.rb +15 -11
- data/spec/delayed/work_queue/parent_process/server_spec.rb +43 -40
- data/spec/delayed/work_queue/parent_process_spec.rb +21 -21
- data/spec/delayed/worker/consul_health_check_spec.rb +22 -22
- data/spec/delayed/worker/health_check_spec.rb +51 -49
- data/spec/delayed/worker_spec.rb +28 -25
- data/spec/gemfiles/52.gemfile +5 -3
- data/spec/gemfiles/52.gemfile.lock +240 -0
- data/spec/gemfiles/60.gemfile +5 -3
- data/spec/gemfiles/60.gemfile.lock +1 -1
- data/spec/gemfiles/61.gemfile +5 -3
- data/spec/sample_jobs.rb +45 -15
- data/spec/shared/delayed_batch.rb +74 -67
- data/spec/shared/delayed_method.rb +143 -102
- data/spec/shared/performable_method.rb +39 -38
- data/spec/shared/shared_backend.rb +534 -441
- data/spec/shared/testing.rb +14 -14
- data/spec/shared/worker.rb +155 -147
- data/spec/shared_jobs_specs.rb +13 -13
- data/spec/spec_helper.rb +43 -40
- metadata +75 -56
- data/lib/delayed/backend/redis/bulk_update.lua +0 -50
- data/lib/delayed/backend/redis/destroy_job.lua +0 -2
- data/lib/delayed/backend/redis/enqueue.lua +0 -29
- data/lib/delayed/backend/redis/fail_job.lua +0 -5
- data/lib/delayed/backend/redis/find_available.lua +0 -3
- data/lib/delayed/backend/redis/functions.rb +0 -59
- data/lib/delayed/backend/redis/get_and_lock_next_available.lua +0 -17
- data/lib/delayed/backend/redis/includes/jobs_common.lua +0 -203
- data/lib/delayed/backend/redis/job.rb +0 -528
- data/lib/delayed/backend/redis/set_running.lua +0 -5
- data/lib/delayed/backend/redis/tickle_strand.lua +0 -2
- data/spec/redis_job_spec.rb +0 -148
@@ -1,7 +1,7 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
3
|
module Delayed
|
4
|
-
|
4
|
+
PerformableMethod = Struct.new(:object, :method, :args, :kwargs, :fail_cb, :permanent_fail_cb, :sender) do # rubocop:disable Lint/StructNewOverride
|
5
5
|
def initialize(object, method, args: [], kwargs: {}, on_failure: nil, on_permanent_failure: nil, sender: nil)
|
6
6
|
raise NoMethodError, "undefined method `#{method}' for #{object.inspect}" unless object.respond_to?(method, true)
|
7
7
|
|
@@ -41,12 +41,10 @@ module Delayed
|
|
41
41
|
else
|
42
42
|
object.send(method, *args, **kwargs)
|
43
43
|
end
|
44
|
+
elsif kwargs.empty?
|
45
|
+
object.public_send(method, *args)
|
44
46
|
else
|
45
|
-
|
46
|
-
object.public_send(method, *args)
|
47
|
-
else
|
48
|
-
object.public_send(method, *args, **kwargs)
|
49
|
-
end
|
47
|
+
object.public_send(method, *args, **kwargs)
|
50
48
|
end
|
51
49
|
end
|
52
50
|
|
@@ -74,7 +72,7 @@ module Delayed
|
|
74
72
|
def full_name
|
75
73
|
obj_name = object.is_a?(ActiveRecord::Base) ? "#{object.class}.find(#{object.id}).#{method}" : display_name
|
76
74
|
kgs = kwargs || {}
|
77
|
-
kwargs_str = kgs.map { |(k, v)| ", #{k}: #{deep_de_ar_ize(v)}"}.join
|
75
|
+
kwargs_str = kgs.map { |(k, v)| ", #{k}: #{deep_de_ar_ize(v)}" }.join
|
78
76
|
"#{obj_name}(#{args.map { |a| deep_de_ar_ize(a) }.join(', ')}#{kwargs_str})"
|
79
77
|
end
|
80
78
|
end
|
data/lib/delayed/periodic.rb
CHANGED
@@ -1,85 +1,86 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
|
-
require
|
3
|
+
require "fugit"
|
4
4
|
|
5
5
|
module Delayed
|
6
|
-
class Periodic
|
7
|
-
|
6
|
+
class Periodic
|
7
|
+
attr_reader :name, :cron
|
8
8
|
|
9
|
-
|
10
|
-
|
11
|
-
|
9
|
+
def encode_with(coder)
|
10
|
+
coder.scalar("!ruby/Delayed::Periodic", @name)
|
11
|
+
end
|
12
12
|
|
13
|
-
|
14
|
-
|
15
|
-
|
13
|
+
cattr_accessor :scheduled, :overrides
|
14
|
+
self.scheduled = {}
|
15
|
+
self.overrides = {}
|
16
16
|
|
17
|
-
|
18
|
-
|
19
|
-
|
20
|
-
|
17
|
+
def self.add_overrides(overrides)
|
18
|
+
overrides.each do |_name, cron_line|
|
19
|
+
# throws error if the line is malformed
|
20
|
+
Fugit.do_parse_cron(cron_line)
|
21
|
+
end
|
22
|
+
self.overrides.merge!(overrides)
|
21
23
|
end
|
22
|
-
self.overrides.merge!(overrides)
|
23
|
-
end
|
24
24
|
|
25
|
-
|
26
|
-
|
27
|
-
cron_line = overrides[job_name] || cron_line
|
28
|
-
self.scheduled[job_name] = self.new(job_name, cron_line, job_args, block)
|
29
|
-
end
|
25
|
+
def self.cron(job_name, cron_line, job_args = {}, &block)
|
26
|
+
raise ArgumentError, "job #{job_name} already scheduled!" if scheduled[job_name]
|
30
27
|
|
31
|
-
|
32
|
-
|
33
|
-
|
34
|
-
# it's fine to just do the audit in-line here without risk of creating duplicates
|
35
|
-
perform_audit!
|
36
|
-
end
|
28
|
+
cron_line = overrides[job_name] || cron_line
|
29
|
+
scheduled[job_name] = new(job_name, cron_line, job_args, block)
|
30
|
+
end
|
37
31
|
|
38
|
-
|
39
|
-
|
40
|
-
|
41
|
-
|
42
|
-
|
32
|
+
def self.audit_queue
|
33
|
+
# we used to queue up a job in a strand here, and perform the audit inside that job
|
34
|
+
# however, now that we're using singletons for scheduling periodic jobs,
|
35
|
+
# it's fine to just do the audit in-line here without risk of creating duplicates
|
36
|
+
perform_audit!
|
37
|
+
end
|
43
38
|
|
44
|
-
|
45
|
-
|
46
|
-
|
47
|
-
|
48
|
-
|
49
|
-
end
|
39
|
+
# make sure all periodic jobs are scheduled for their next run in the job queue
|
40
|
+
# this auditing should run on the strand
|
41
|
+
def self.perform_audit!
|
42
|
+
scheduled.each { |_name, periodic| periodic.enqueue }
|
43
|
+
end
|
50
44
|
|
51
|
-
|
52
|
-
|
53
|
-
|
45
|
+
def initialize(name, cron_line, job_args, block)
|
46
|
+
@name = name
|
47
|
+
@cron = Fugit.do_parse_cron(cron_line)
|
48
|
+
@job_args = { priority: Delayed::LOW_PRIORITY }.merge(job_args.symbolize_keys)
|
49
|
+
@block = block
|
50
|
+
end
|
54
51
|
|
55
|
-
|
56
|
-
|
57
|
-
|
58
|
-
run_at: @cron.next_time(Delayed::Periodic.now).utc.to_time,
|
59
|
-
singleton: tag,
|
60
|
-
on_conflict: :patient
|
61
|
-
}
|
62
|
-
@job_args.merge(inferred_args)
|
63
|
-
end
|
52
|
+
def enqueue
|
53
|
+
Delayed::Job.enqueue(self, **enqueue_args)
|
54
|
+
end
|
64
55
|
|
65
|
-
|
66
|
-
|
67
|
-
|
68
|
-
|
69
|
-
|
70
|
-
|
71
|
-
|
72
|
-
|
56
|
+
def enqueue_args
|
57
|
+
inferred_args = {
|
58
|
+
max_attempts: 1,
|
59
|
+
run_at: @cron.next_time(Delayed::Periodic.now).utc.to_time,
|
60
|
+
singleton: tag,
|
61
|
+
on_conflict: :patient
|
62
|
+
}
|
63
|
+
@job_args.merge(inferred_args)
|
73
64
|
end
|
74
|
-
end
|
75
65
|
|
76
|
-
|
77
|
-
|
78
|
-
|
79
|
-
|
66
|
+
def perform
|
67
|
+
@block.call
|
68
|
+
ensure
|
69
|
+
begin
|
70
|
+
enqueue
|
71
|
+
rescue
|
72
|
+
# double fail! the auditor will have to catch this.
|
73
|
+
Rails.logger.error "Failure enqueueing periodic job! #{@name} #{$!.inspect}"
|
74
|
+
end
|
75
|
+
end
|
76
|
+
|
77
|
+
def tag
|
78
|
+
"periodic: #{@name}"
|
79
|
+
end
|
80
|
+
alias display_name tag
|
80
81
|
|
81
|
-
|
82
|
-
|
82
|
+
def self.now
|
83
|
+
Time.zone.now
|
84
|
+
end
|
83
85
|
end
|
84
86
|
end
|
85
|
-
end
|
data/lib/delayed/plugin.rb
CHANGED
@@ -1,6 +1,6 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
|
-
require
|
3
|
+
require "active_support/core_ext/class/attribute"
|
4
4
|
|
5
5
|
module Delayed
|
6
6
|
class Plugin
|
@@ -11,9 +11,7 @@ module Delayed
|
|
11
11
|
end
|
12
12
|
|
13
13
|
def self.inject!
|
14
|
-
unless @injected
|
15
|
-
self.callback_block.call(Delayed::Worker.lifecycle) if self.callback_block
|
16
|
-
end
|
14
|
+
callback_block&.call(Delayed::Worker.lifecycle) unless @injected
|
17
15
|
@injected = true
|
18
16
|
end
|
19
17
|
|
data/lib/delayed/pool.rb
CHANGED
@@ -1,248 +1,254 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
3
|
module Delayed
|
4
|
-
class Pool
|
5
|
-
|
4
|
+
class Pool
|
5
|
+
include Delayed::Logging
|
6
6
|
|
7
|
-
|
8
|
-
|
7
|
+
mattr_accessor :on_fork
|
8
|
+
self.on_fork = -> {}
|
9
9
|
|
10
|
-
|
11
|
-
|
10
|
+
SIGNALS = %i[INT TERM QUIT].freeze
|
11
|
+
POOL_SLEEP_PERIOD = 5
|
12
12
|
|
13
|
-
|
13
|
+
attr_reader :workers
|
14
14
|
|
15
|
-
|
16
|
-
|
17
|
-
|
18
|
-
|
19
|
-
|
15
|
+
def initialize(*args)
|
16
|
+
if args.first.is_a?(Hash)
|
17
|
+
@config = args.first
|
18
|
+
else
|
19
|
+
warn "Calling Delayed::Pool.new directly is deprecated. Use `Delayed::CLI.new.run()` instead."
|
20
|
+
end
|
21
|
+
@workers = {}
|
22
|
+
@signal_queue = []
|
23
|
+
@self_pipe = IO.pipe
|
20
24
|
end
|
21
|
-
@workers = {}
|
22
|
-
@signal_queue = []
|
23
|
-
@self_pipe = IO.pipe
|
24
|
-
end
|
25
25
|
|
26
|
-
|
27
|
-
|
28
|
-
|
29
|
-
|
26
|
+
def run
|
27
|
+
warn "Delayed::Pool#run is deprecated and will be removed. Use `Delayed::CLI.new.run()` instead."
|
28
|
+
Delayed::CLI.new.run
|
29
|
+
end
|
30
30
|
|
31
|
-
|
32
|
-
|
33
|
-
|
34
|
-
|
35
|
-
|
36
|
-
|
37
|
-
|
31
|
+
def start
|
32
|
+
say "Started job master", :info
|
33
|
+
SIGNALS.each do |sig|
|
34
|
+
trap(sig) do
|
35
|
+
@signal_queue << sig
|
36
|
+
wake_up
|
37
|
+
end
|
38
|
+
end
|
39
|
+
$0 = procname
|
40
|
+
# fork to handle unlocking (to prevent polluting the parent with worker objects)
|
41
|
+
unlock_pid = fork_with_reconnects do
|
42
|
+
unlock_orphaned_jobs
|
43
|
+
end
|
44
|
+
Process.wait unlock_pid
|
45
|
+
|
46
|
+
spawn_periodic_auditor
|
47
|
+
spawn_abandoned_job_cleanup
|
48
|
+
spawn_all_workers
|
49
|
+
say "Workers spawned"
|
50
|
+
join
|
51
|
+
say "Shutting down"
|
52
|
+
stop
|
53
|
+
reap_all_children
|
54
|
+
rescue Exception => e # rubocop:disable Lint/RescueException
|
55
|
+
say "Job master died with error: #{e.inspect}\n#{e.backtrace.join("\n")}", :fatal
|
56
|
+
raise
|
38
57
|
end
|
39
|
-
Process.wait unlock_pid
|
40
|
-
|
41
|
-
spawn_periodic_auditor
|
42
|
-
spawn_abandoned_job_cleanup
|
43
|
-
spawn_all_workers
|
44
|
-
say "Workers spawned"
|
45
|
-
join
|
46
|
-
say "Shutting down"
|
47
|
-
stop
|
48
|
-
reap_all_children
|
49
|
-
rescue Exception => e
|
50
|
-
say "Job master died with error: #{e.inspect}\n#{e.backtrace.join("\n")}", :fatal
|
51
|
-
raise
|
52
|
-
end
|
53
58
|
|
54
|
-
|
59
|
+
protected
|
55
60
|
|
56
|
-
|
57
|
-
|
58
|
-
|
61
|
+
def procname
|
62
|
+
"delayed_jobs_pool#{Settings.pool_procname_suffix}"
|
63
|
+
end
|
59
64
|
|
60
|
-
|
61
|
-
|
65
|
+
def unlock_orphaned_jobs(_worker = nil, pid = nil)
|
66
|
+
return if Settings.disable_automatic_orphan_unlocking
|
62
67
|
|
63
|
-
|
64
|
-
|
65
|
-
|
66
|
-
|
68
|
+
unlocked_jobs = Delayed::Job.unlock_orphaned_jobs(pid)
|
69
|
+
say "Unlocked #{unlocked_jobs} orphaned jobs" if unlocked_jobs.positive?
|
70
|
+
ActiveRecord::Base.connection_handler.clear_all_connections! unless Rails.env.test?
|
71
|
+
end
|
67
72
|
|
68
|
-
|
69
|
-
|
73
|
+
def spawn_all_workers
|
74
|
+
ActiveRecord::Base.connection_handler.clear_all_connections!
|
70
75
|
|
71
|
-
|
72
|
-
|
73
|
-
|
74
|
-
|
76
|
+
if @config[:work_queue] == "parent_process"
|
77
|
+
@work_queue = WorkQueue::ParentProcess.new
|
78
|
+
spawn_work_queue
|
79
|
+
end
|
75
80
|
|
76
|
-
|
77
|
-
|
81
|
+
@config[:workers].each do |worker_config|
|
82
|
+
(worker_config[:workers] || 1).times { spawn_worker(worker_config) }
|
83
|
+
end
|
78
84
|
end
|
79
|
-
end
|
80
85
|
|
81
|
-
|
82
|
-
|
83
|
-
|
84
|
-
|
85
|
-
|
86
|
+
def spawn_work_queue
|
87
|
+
parent_pid = Process.pid
|
88
|
+
pid = fork_with_reconnects do
|
89
|
+
$0 = "delayed_jobs_work_queue#{Settings.pool_procname_suffix}"
|
90
|
+
@work_queue.server(parent_pid: parent_pid).run
|
91
|
+
end
|
92
|
+
workers[pid] = :work_queue
|
86
93
|
end
|
87
|
-
workers[pid] = :work_queue
|
88
|
-
end
|
89
94
|
|
90
|
-
|
91
|
-
|
92
|
-
|
93
|
-
else
|
95
|
+
def spawn_worker(worker_config)
|
96
|
+
return if worker_config[:periodic] # backwards compat
|
97
|
+
|
94
98
|
worker_config[:parent_pid] = Process.pid
|
95
99
|
worker_config[:work_queue] = @work_queue.client if @work_queue
|
96
100
|
worker = Delayed::Worker.new(worker_config)
|
97
|
-
end
|
98
101
|
|
99
|
-
|
100
|
-
|
102
|
+
pid = fork_with_reconnects do
|
103
|
+
worker.start
|
104
|
+
end
|
105
|
+
workers[pid] = worker
|
101
106
|
end
|
102
|
-
workers[pid] = worker
|
103
|
-
end
|
104
107
|
|
105
|
-
|
106
|
-
|
107
|
-
|
108
|
-
|
109
|
-
|
110
|
-
|
111
|
-
|
112
|
-
|
108
|
+
# child processes need to reconnect so they don't accidentally share redis or
|
109
|
+
# db connections with the parent
|
110
|
+
def fork_with_reconnects
|
111
|
+
fork do
|
112
|
+
@self_pipe.each(&:close) # sub-processes don't need to wake us up; keep their FDs clean
|
113
|
+
Pool.on_fork.call
|
114
|
+
Delayed::Job.reconnect!
|
115
|
+
yield
|
116
|
+
end
|
113
117
|
end
|
114
|
-
end
|
115
118
|
|
116
|
-
|
117
|
-
|
118
|
-
|
119
|
-
|
120
|
-
|
121
|
-
|
122
|
-
|
123
|
-
|
124
|
-
|
125
|
-
|
126
|
-
|
127
|
-
|
128
|
-
|
129
|
-
|
130
|
-
|
119
|
+
def spawn_abandoned_job_cleanup
|
120
|
+
return if Settings.disable_abandoned_job_cleanup
|
121
|
+
|
122
|
+
cleanup_interval_in_minutes = 60
|
123
|
+
@abandoned_cleanup_thread = Thread.new do
|
124
|
+
# every hour (staggered by process)
|
125
|
+
# check for dead jobs and cull them.
|
126
|
+
# Will actually be more often based on the
|
127
|
+
# number of worker nodes in the pool. This will actually
|
128
|
+
# be a max of N times per hour where N is the number of workers,
|
129
|
+
# but they won't overrun each other because the health check
|
130
|
+
# takes an advisory lock internally
|
131
|
+
sleep(rand(cleanup_interval_in_minutes * 60))
|
132
|
+
loop do
|
133
|
+
schedule_abandoned_job_cleanup
|
134
|
+
sleep(cleanup_interval_in_minutes * 60)
|
135
|
+
end
|
131
136
|
end
|
132
137
|
end
|
133
|
-
end
|
134
138
|
|
135
|
-
|
136
|
-
|
137
|
-
|
138
|
-
|
139
|
-
|
139
|
+
def schedule_abandoned_job_cleanup
|
140
|
+
pid = fork_with_reconnects do
|
141
|
+
# we want to avoid db connections in the main pool process
|
142
|
+
$0 = "delayed_abandoned_job_cleanup"
|
143
|
+
Delayed::Worker::HealthCheck.reschedule_abandoned_jobs
|
144
|
+
end
|
145
|
+
workers[pid] = :abandoned_job_cleanup
|
140
146
|
end
|
141
|
-
workers[pid] = :abandoned_job_cleanup
|
142
|
-
end
|
143
147
|
|
144
|
-
|
145
|
-
|
148
|
+
def spawn_periodic_auditor
|
149
|
+
return if Settings.disable_periodic_jobs
|
146
150
|
|
147
|
-
|
148
|
-
|
149
|
-
schedule_periodic_audit
|
150
|
-
# initial sleep is randomized, for some staggering in the audit calls
|
151
|
-
# since job processors are usually all restarted at the same time
|
152
|
-
sleep(rand(15 * 60))
|
153
|
-
loop do
|
151
|
+
@periodic_thread = Thread.new do
|
152
|
+
# schedule the initial audit immediately on startup
|
154
153
|
schedule_periodic_audit
|
155
|
-
sleep
|
154
|
+
# initial sleep is randomized, for some staggering in the audit calls
|
155
|
+
# since job processors are usually all restarted at the same time
|
156
|
+
sleep(rand(15 * 60))
|
157
|
+
loop do
|
158
|
+
schedule_periodic_audit
|
159
|
+
sleep(15 * 60)
|
160
|
+
end
|
156
161
|
end
|
157
162
|
end
|
158
|
-
end
|
159
163
|
|
160
|
-
|
161
|
-
|
162
|
-
|
163
|
-
|
164
|
-
|
164
|
+
def schedule_periodic_audit
|
165
|
+
pid = fork_with_reconnects do
|
166
|
+
# we want to avoid db connections in the main pool process
|
167
|
+
$0 = "delayed_periodic_audit_scheduler"
|
168
|
+
Delayed::Periodic.audit_queue
|
169
|
+
end
|
170
|
+
workers[pid] = :periodic_audit
|
165
171
|
end
|
166
|
-
workers[pid] = :periodic_audit
|
167
|
-
end
|
168
172
|
|
169
|
-
|
170
|
-
|
171
|
-
|
172
|
-
|
173
|
-
|
174
|
-
|
175
|
-
|
176
|
-
|
177
|
-
|
178
|
-
|
179
|
-
|
180
|
-
|
181
|
-
|
173
|
+
def join
|
174
|
+
loop do
|
175
|
+
maintain_children
|
176
|
+
case sig = @signal_queue.shift
|
177
|
+
when nil
|
178
|
+
pool_sleep
|
179
|
+
when :QUIT
|
180
|
+
break
|
181
|
+
when :TERM, :INT
|
182
|
+
stop(graceful: false) if Settings.kill_workers_on_exit
|
183
|
+
break
|
184
|
+
else
|
185
|
+
logger.warn("Unexpected signal received: #{sig}")
|
186
|
+
end
|
182
187
|
end
|
183
188
|
end
|
184
|
-
end
|
185
189
|
|
186
|
-
|
187
|
-
|
188
|
-
|
189
|
-
|
190
|
+
def pool_sleep
|
191
|
+
IO.select([@self_pipe[0]], nil, nil, POOL_SLEEP_PERIOD)
|
192
|
+
@self_pipe[0].read_nonblock(11, exception: false)
|
193
|
+
end
|
190
194
|
|
191
|
-
|
192
|
-
|
193
|
-
|
194
|
-
|
195
|
-
|
195
|
+
def stop(graceful: true, timeout: Settings.slow_exit_timeout)
|
196
|
+
signal_for_children = graceful ? :QUIT : :TERM
|
197
|
+
if Settings.kill_workers_on_exit
|
198
|
+
limit = Time.zone.now + timeout
|
199
|
+
until @workers.empty? || Time.zone.now >= limit
|
200
|
+
signal_all_children(signal_for_children)
|
201
|
+
# Give our children some time to process the signal before checking if
|
202
|
+
# they've exited
|
203
|
+
sleep(0.5)
|
204
|
+
reap_all_children.each { |pid| @workers.delete(pid) }
|
205
|
+
end
|
206
|
+
|
207
|
+
# We really want to give the workers every oportunity to clean up after
|
208
|
+
# themselves before murdering them.
|
209
|
+
stop(graceful: false, timeout: 2) if graceful
|
210
|
+
signal_all_children(:KILL)
|
211
|
+
else
|
196
212
|
signal_all_children(signal_for_children)
|
197
|
-
# Give our children some time to process the signal before checking if
|
198
|
-
# they've exited
|
199
|
-
sleep(0.5)
|
200
|
-
reap_all_children.each { |pid| @workers.delete(pid) }
|
201
213
|
end
|
214
|
+
end
|
202
215
|
|
203
|
-
|
204
|
-
|
205
|
-
stop(graceful: false, timeout: 2) if graceful
|
206
|
-
signal_all_children(:KILL)
|
207
|
-
else
|
208
|
-
signal_all_children(signal_for_children)
|
216
|
+
def signal_all_children(signal)
|
217
|
+
workers.each_key { |pid| signal_child(signal, pid) }
|
209
218
|
end
|
210
|
-
end
|
211
219
|
|
212
|
-
|
213
|
-
|
214
|
-
|
220
|
+
def signal_child(signal, pid)
|
221
|
+
Process.kill(signal, pid)
|
222
|
+
rescue Erron::ESRCH
|
223
|
+
workers.delete(pid)
|
224
|
+
end
|
215
225
|
|
216
|
-
|
217
|
-
|
218
|
-
|
219
|
-
|
220
|
-
|
226
|
+
# Respawn all children that have exited since we last checked
|
227
|
+
def maintain_children
|
228
|
+
reap_all_children.each do |pid|
|
229
|
+
respawn_child(pid)
|
230
|
+
end
|
231
|
+
end
|
232
|
+
|
233
|
+
# Reaps processes that have exited or just returns if none have exited
|
234
|
+
#
|
235
|
+
# @return Array An array of child pids that have exited
|
236
|
+
def reap_all_children
|
237
|
+
exited_pids = []
|
238
|
+
loop do
|
239
|
+
pid = Process.wait(-1, Process::WNOHANG)
|
240
|
+
break unless pid
|
221
241
|
|
222
|
-
|
223
|
-
|
224
|
-
|
225
|
-
|
242
|
+
exited_pids << pid
|
243
|
+
rescue Errno::ECHILD
|
244
|
+
break
|
245
|
+
end
|
246
|
+
exited_pids
|
226
247
|
end
|
227
|
-
end
|
228
248
|
|
229
|
-
|
230
|
-
|
231
|
-
# @return Array An array of child pids that have exited
|
232
|
-
def reap_all_children
|
233
|
-
exited_pids = []
|
234
|
-
begin
|
235
|
-
pid = Process.wait(-1, Process::WNOHANG)
|
236
|
-
break unless pid
|
237
|
-
exited_pids << pid
|
238
|
-
rescue Errno::ECHILD
|
239
|
-
break
|
240
|
-
end while true # I hate doing loops this way but it's the only way to make the rescue work
|
241
|
-
exited_pids
|
242
|
-
end
|
249
|
+
def respawn_child(child)
|
250
|
+
return unless workers.include?(child)
|
243
251
|
|
244
|
-
def respawn_child(child)
|
245
|
-
if workers.include?(child)
|
246
252
|
worker = workers.delete(child)
|
247
253
|
case worker
|
248
254
|
when :periodic_audit
|
@@ -262,10 +268,9 @@ class Pool
|
|
262
268
|
spawn_worker(worker.config)
|
263
269
|
end
|
264
270
|
end
|
265
|
-
end
|
266
271
|
|
267
|
-
|
268
|
-
|
272
|
+
def wake_up
|
273
|
+
@self_pipe[1].write_nonblock(".", exception: false)
|
274
|
+
end
|
269
275
|
end
|
270
276
|
end
|
271
|
-
end
|