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
|
# frozen_string_literal: true
module Gitlab
module Database
module Migrations
module BackgroundMigrationHelpers
BACKGROUND_MIGRATION_BATCH_SIZE = 1_000 # Number of rows to process per job
BACKGROUND_MIGRATION_JOB_BUFFER_SIZE = 1_000 # Number of jobs to bulk queue at a time
# Bulk queues background migration jobs for an entire table, batched by ID range.
# "Bulk" meaning many jobs will be pushed at a time for efficiency.
# If you need a delay interval per job, then use `queue_background_migration_jobs_by_range_at_intervals`.
#
# model_class - The table being iterated over
# job_class_name - The background migration job class as a string
# batch_size - The maximum number of rows per job
#
# Example:
#
# class Route < ActiveRecord::Base
# include EachBatch
# self.table_name = 'routes'
# end
#
# bulk_queue_background_migration_jobs_by_range(Route, 'ProcessRoutes')
#
# Where the model_class includes EachBatch, and the background migration exists:
#
# class Gitlab::BackgroundMigration::ProcessRoutes
# def perform(start_id, end_id)
# # do something
# end
# end
def bulk_queue_background_migration_jobs_by_range(model_class, job_class_name, batch_size: BACKGROUND_MIGRATION_BATCH_SIZE)
raise "#{model_class} does not have an ID to use for batch ranges" unless model_class.column_names.include?('id')
jobs = []
table_name = model_class.quoted_table_name
model_class.each_batch(of: batch_size) do |relation|
start_id, end_id = relation.pluck("MIN(#{table_name}.id)", "MAX(#{table_name}.id)").first
if jobs.length >= BACKGROUND_MIGRATION_JOB_BUFFER_SIZE
# Note: This code path generally only helps with many millions of rows
# We push multiple jobs at a time to reduce the time spent in
# Sidekiq/Redis operations. We're using this buffer based approach so we
# don't need to run additional queries for every range.
bulk_migrate_async(jobs)
jobs.clear
end
jobs << [job_class_name, [start_id, end_id]]
end
bulk_migrate_async(jobs) unless jobs.empty?
end
# Queues background migration jobs for an entire table in batches.
# The default batching column used is the standard primary key `id`.
# Each job is scheduled with a `delay_interval` in between.
# If you use a small interval, then some jobs may run at the same time.
#
# model_class - The table or relation being iterated over
# job_class_name - The background migration job class as a string
# delay_interval - The duration between each job's scheduled time (must respond to `to_f`)
# batch_size - The maximum number of rows per job
# other_arguments - Other arguments to send to the job
# track_jobs - When this flag is set, creates a record in the background_migration_jobs table for each job that
# is scheduled to be run. These records can be used to trace execution of the background job, but there is no
# builtin support to manage that automatically at this time. You should only set this flag if you are aware of
# how it works, and intend to manually cleanup the database records in your background job.
# primary_column_name - The name of the primary key column if the primary key is not `id`
#
# *Returns the final migration delay*
#
# Example:
#
# class Route < ActiveRecord::Base
# include EachBatch
# self.table_name = 'routes'
# end
#
# queue_background_migration_jobs_by_range_at_intervals(Route, 'ProcessRoutes', 1.minute)
#
# Where the model_class includes EachBatch, and the background migration exists:
#
# class Gitlab::BackgroundMigration::ProcessRoutes
# def perform(start_id, end_id)
# # do something
# end
# end
def queue_background_migration_jobs_by_range_at_intervals(model_class, job_class_name, delay_interval, batch_size: BACKGROUND_MIGRATION_BATCH_SIZE, other_job_arguments: [], initial_delay: 0, track_jobs: false, primary_column_name: :id)
raise "#{model_class} does not have an ID column of #{primary_column_name} to use for batch ranges" unless model_class.column_names.include?(primary_column_name.to_s)
raise "#{primary_column_name} is not an integer column" unless model_class.columns_hash[primary_column_name.to_s].type == :integer
# To not overload the worker too much we enforce a minimum interval both
# when scheduling and performing jobs.
if delay_interval < BackgroundMigrationWorker.minimum_interval
delay_interval = BackgroundMigrationWorker.minimum_interval
end
final_delay = 0
batch_counter = 0
model_class.each_batch(of: batch_size) do |relation, index|
start_id, end_id = relation.pluck(Arel.sql("MIN(#{primary_column_name}), MAX(#{primary_column_name})")).first
# `BackgroundMigrationWorker.bulk_perform_in` schedules all jobs for
# the same time, which is not helpful in most cases where we wish to
# spread the work over time.
final_delay = initial_delay + delay_interval * index
full_job_arguments = [start_id, end_id] + other_job_arguments
track_in_database(job_class_name, full_job_arguments) if track_jobs
migrate_in(final_delay, job_class_name, full_job_arguments)
batch_counter += 1
end
duration = initial_delay + delay_interval * batch_counter
say <<~SAY
Scheduled #{batch_counter} #{job_class_name} jobs with a maximum of #{batch_size} records per batch and an interval of #{delay_interval} seconds.
The migration is expected to take at least #{duration} seconds. Expect all jobs to have completed after #{Time.zone.now + duration}."
SAY
final_delay
end
def perform_background_migration_inline?
Rails.env.test? || Rails.env.development?
end
def migrate_async(*args)
with_migration_context do
BackgroundMigrationWorker.perform_async(*args)
end
end
def migrate_in(*args)
with_migration_context do
BackgroundMigrationWorker.perform_in(*args)
end
end
def bulk_migrate_in(*args)
with_migration_context do
BackgroundMigrationWorker.bulk_perform_in(*args)
end
end
def bulk_migrate_async(*args)
with_migration_context do
BackgroundMigrationWorker.bulk_perform_async(*args)
end
end
def with_migration_context(&block)
Gitlab::ApplicationContext.with_context(caller_id: self.class.to_s, &block)
end
private
def track_in_database(class_name, arguments)
Gitlab::Database::BackgroundMigrationJob.create!(class_name: class_name, arguments: arguments)
end
end
end
end
end
|