1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
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
207
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
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
|
# frozen_string_literal: true
require_relative 'interrupted_set'
module Sidekiq
class BaseReliableFetch
DEFAULT_CLEANUP_INTERVAL = 60 * 60 # 1 hour
HEARTBEAT_INTERVAL = 20 # seconds
HEARTBEAT_LIFESPAN = 60 # seconds
HEARTBEAT_RETRY_DELAY = 1 # seconds
WORKING_QUEUE_PREFIX = 'working'
# Defines how often we try to take a lease to not flood our
# Redis server with SET requests
DEFAULT_LEASE_INTERVAL = 2 * 60 # seconds
LEASE_KEY = 'reliable-fetcher-cleanup-lock'
# Defines the COUNT parameter that will be passed to Redis SCAN command
SCAN_COUNT = 1000
# How much time a job can be interrupted
DEFAULT_MAX_RETRIES_AFTER_INTERRUPTION = 3
# Regexes for matching working queue keys
WORKING_QUEUE_REGEX = /#{WORKING_QUEUE_PREFIX}:(queue:.*):([^:]*:[0-9]*:[0-9a-f]*)\z/.freeze
LEGACY_WORKING_QUEUE_REGEX = /#{WORKING_QUEUE_PREFIX}:(queue:.*):([^:]*:[0-9]*)\z/.freeze
UnitOfWork = Struct.new(:queue, :job) do
def acknowledge
Sidekiq.redis { |conn| conn.lrem(Sidekiq::BaseReliableFetch.working_queue_name(queue), 1, job) }
end
def queue_name
queue.sub(/.*queue:/, '')
end
def requeue
Sidekiq.redis do |conn|
conn.multi do |multi|
multi.lpush(queue, job)
multi.lrem(Sidekiq::BaseReliableFetch.working_queue_name(queue), 1, job)
end
end
end
end
def self.setup_reliable_fetch!(config)
config = config.options unless config.respond_to?(:[])
fetch_strategy = if config[:semi_reliable_fetch]
Sidekiq::SemiReliableFetch
else
Sidekiq::ReliableFetch
end
config[:fetch_class] = fetch_strategy
Sidekiq.logger.info('GitLab reliable fetch activated!')
# Set the heartbeat immediately to prevent a race condition where
# worker_dead? returns true in another thread. `start_heartbeat_thread`
# isn't guaranteed to have run before Sidekiq attempts to fetch jobs.
heartbeat
start_heartbeat_thread
end
def self.start_heartbeat_thread
Thread.new do
loop do
begin
heartbeat
sleep HEARTBEAT_INTERVAL
rescue => e
Sidekiq.logger.error("Heartbeat thread error: #{e.message}")
sleep HEARTBEAT_RETRY_DELAY
end
end
end
end
def self.hostname
Socket.gethostname
end
def self.process_nonce
@@process_nonce ||= SecureRandom.hex(6)
end
def self.identity
@@identity ||= "#{hostname}:#{$$}:#{process_nonce}"
end
def self.heartbeat
Sidekiq.redis do |conn|
conn.set(heartbeat_key(identity), 1, ex: HEARTBEAT_LIFESPAN)
end
Sidekiq.logger.debug("Heartbeat for #{identity}")
end
def self.worker_dead?(identity, conn)
!conn.get(heartbeat_key(identity))
end
def self.heartbeat_key(identity)
"reliable-fetcher-heartbeat-#{identity.gsub(':', '-')}"
end
def self.working_queue_name(queue)
"#{WORKING_QUEUE_PREFIX}:#{queue}:#{identity}"
end
attr_reader :cleanup_interval, :last_try_to_take_lease_at, :lease_interval,
:queues, :use_semi_reliable_fetch,
:strictly_ordered_queues, :config
def initialize(capsule)
raise ArgumentError, 'missing queue list' unless capsule.config.queues
@config = capsule.config
@interrupted_set = Sidekiq::InterruptedSet.new
@cleanup_interval = config.fetch(:cleanup_interval, DEFAULT_CLEANUP_INTERVAL)
@lease_interval = config.fetch(:lease_interval, DEFAULT_LEASE_INTERVAL)
@last_try_to_take_lease_at = 0
@strictly_ordered_queues = !!config[:strict]
@queues = config.queues.map { |q| "queue:#{q}" }
end
def retrieve_work
clean_working_queues! if take_lease
retrieve_unit_of_work
end
def retrieve_unit_of_work
raise NotImplementedError,
"#{self.class} does not implement #{__method__}"
end
def bulk_requeue(inprogress)
return if inprogress.empty?
Sidekiq.redis do |conn|
inprogress.each do |unit_of_work|
conn.multi do |multi|
preprocess_interrupted_job(unit_of_work.job, unit_of_work.queue, multi)
multi.lrem(self.class.working_queue_name(unit_of_work.queue), 1, unit_of_work.job)
end
end
end
rescue => e
Sidekiq.logger.warn("Failed to requeue #{inprogress.size} jobs: #{e.message}")
end
private
def preprocess_interrupted_job(job, queue, conn = nil)
msg = Sidekiq.load_json(job)
msg['interrupted_count'] = msg['interrupted_count'].to_i + 1
if interruption_exhausted?(msg)
send_to_quarantine(msg, conn)
else
requeue_job(queue, msg, conn)
end
end
# If you want this method to be run in a scope of multi connection
# you need to pass it
def requeue_job(queue, msg, conn)
with_connection(conn) do |conn|
conn.lpush(queue, Sidekiq.dump_json(msg))
end
Sidekiq.logger.info(
message: "Pushed job #{msg['jid']} back to queue #{queue}",
jid: msg['jid'],
class: msg['class'],
queue: queue
)
end
def extract_queue_and_identity(key)
# New identity format is "{hostname}:{pid}:{randomhex}
# Old identity format is "{hostname}:{pid}"
# Queue names may also have colons (namespaced).
# Expressing this in a single regex is unreadable
# Test the newer expected format first, only checking the older if necessary
original_queue, identity = key.scan(WORKING_QUEUE_REGEX).flatten
return original_queue, identity unless original_queue.nil? || identity.nil?
key.scan(LEGACY_WORKING_QUEUE_REGEX).flatten
end
# Detect "old" jobs and requeue them because the worker they were assigned
# to probably failed miserably.
def clean_working_queues!
Sidekiq.logger.info('Cleaning working queues')
Sidekiq.redis do |conn|
conn.scan(match: "#{WORKING_QUEUE_PREFIX}:queue:*", count: SCAN_COUNT) do |key|
original_queue, identity = extract_queue_and_identity(key)
next if original_queue.nil? || identity.nil?
clean_working_queue!(original_queue, key) if self.class.worker_dead?(identity, conn)
end
end
end
def clean_working_queue!(original_queue, working_queue)
Sidekiq.redis do |conn|
while job = conn.rpop(working_queue)
preprocess_interrupted_job(job, original_queue)
end
end
end
def interruption_exhausted?(msg)
return false if max_retries_after_interruption(msg['class']) < 0
msg['interrupted_count'].to_i >= max_retries_after_interruption(msg['class'])
end
def max_retries_after_interruption(worker_class)
max_retries_after_interruption = nil
max_retries_after_interruption ||= begin
Object.const_get(worker_class).sidekiq_options['max_retries_after_interruption']
rescue NameError
end
max_retries_after_interruption ||= config[:max_retries_after_interruption]
max_retries_after_interruption ||= DEFAULT_MAX_RETRIES_AFTER_INTERRUPTION
max_retries_after_interruption
end
def send_to_quarantine(msg, multi_connection = nil)
Sidekiq.logger.warn(
class: msg['class'],
jid: msg['jid'],
message: %(Reliable Fetcher: adding dead #{msg['class']} job #{msg['jid']} to interrupted queue)
)
job = Sidekiq.dump_json(msg)
@interrupted_set.put(job, connection: multi_connection)
end
# Yield block with an existing connection or creates another one
def with_connection(conn)
return yield(conn) if conn
Sidekiq.redis { |redis_conn| yield(redis_conn) }
end
def take_lease
return unless allowed_to_take_a_lease?
@last_try_to_take_lease_at = Time.now.to_f
Sidekiq.redis do |conn|
conn.set(LEASE_KEY, 1, 'nx', 'ex', cleanup_interval)
end
end
def allowed_to_take_a_lease?
Time.now.to_f - last_try_to_take_lease_at > lease_interval
end
end
end
|