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
|
# frozen_string_literal: true
module Gitlab
module Database
module LoadBalancing
class SidekiqServerMiddleware
JobReplicaNotUpToDate = Class.new(StandardError)
def call(worker, job, _queue)
worker_class = worker.class
strategy = select_load_balancing_strategy(worker_class, job)
job['load_balancing_strategy'] = strategy.to_s
if use_primary?(strategy)
Session.current.use_primary!
elsif strategy == :retry
raise JobReplicaNotUpToDate, "Sidekiq job #{worker_class} JID-#{job['jid']} couldn't use the replica."\
" Replica was not up to date."
else
# this means we selected an up-to-date replica, but there is nothing to do in this case.
end
yield
ensure
clear
end
private
def clear
release_hosts
Session.clear_session
end
def use_primary?(strategy)
strategy.start_with?('primary')
end
def select_load_balancing_strategy(worker_class, job)
return :primary unless load_balancing_available?(worker_class)
wal_locations = get_wal_locations(job)
return :primary_no_wal unless wal_locations
if all_databases_has_replica_caught_up?(wal_locations)
# Happy case: we can read from a replica.
retried_before?(worker_class, job) ? :replica_retried : :replica
elsif can_retry?(worker_class, job)
# Optimistic case: The worker allows retries and we have retries left.
:retry
else
# Sad case: we need to fall back to the primary.
:primary
end
end
def get_wal_locations(job)
job['dedup_wal_locations'] || job['wal_locations'] || legacy_wal_location(job)
end
# Already scheduled jobs could still contain legacy database write location.
# TODO: remove this in the next iteration
# https://gitlab.com/gitlab-org/gitlab/-/issues/338213
def legacy_wal_location(job)
wal_location = job['database_write_location'] || job['database_replica_location']
{ Gitlab::Database::MAIN_DATABASE_NAME.to_sym => wal_location } if wal_location
end
def load_balancing_available?(worker_class)
worker_class.include?(::ApplicationWorker) &&
worker_class.utilizes_load_balancing_capabilities? &&
worker_class.get_data_consistency_feature_flag_enabled?
end
def can_retry?(worker_class, job)
worker_class.get_data_consistency == :delayed && not_yet_retried?(job)
end
def retried_before?(worker_class, job)
worker_class.get_data_consistency == :delayed && !not_yet_retried?(job)
end
def not_yet_retried?(job)
# if `retry_count` is `nil` it indicates that this job was never retried
# the `0` indicates that this is a first retry
job['retry_count'].nil?
end
def all_databases_has_replica_caught_up?(wal_locations)
wal_locations.all? do |_config_name, location|
# Once we add support for multiple databases to our load balancer, we would use something like this:
# Gitlab::Database::DATABASES[config_name].load_balancer.select_up_to_date_host(location)
load_balancer.select_up_to_date_host(location)
end
end
def release_hosts
# Once we add support for multiple databases to our load balancer, we would use something like this:
# connection.load_balancer.primary_write_location
#
# Gitlab::Database::DATABASES.values.each do |connection|
# connection.load_balancer.release_host
# end
load_balancer.release_host
end
def load_balancer
LoadBalancing.proxy.load_balancer
end
end
end
end
end
|