blob: 1e43e3dd79074e02f3cb5f6df1ff62e0c4129d98 (
plain)
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
|
# frozen_string_literal: true
# See http://doc.gitlab.com/ce/development/migration_style_guide.html
# for more information on how to write migrations for GitLab.
class PopulateMrMetricsWithEventsData < ActiveRecord::Migration[4.2]
include Gitlab::Database::MigrationHelpers
DOWNTIME = false
BATCH_SIZE = 10_000
MIGRATION = 'PopulateMergeRequestMetricsWithEventsDataImproved'
PREVIOUS_MIGRATION = 'PopulateMergeRequestMetricsWithEventsData'
disable_ddl_transaction!
def up
# Perform any ongoing background migration that might still be running from
# previous try (see https://gitlab.com/gitlab-org/gitlab-ce/issues/47676).
Gitlab::BackgroundMigration.steal(PREVIOUS_MIGRATION)
say 'Scheduling `PopulateMergeRequestMetricsWithEventsData` jobs'
# It will update around 4_000_000 records in batches of 10_000 merge
# requests (running between 5 minutes) and should take around 53 hours to complete.
# Apparently, production PostgreSQL is able to vacuum 10k-20k dead_tuples
# per minute. So this should give us enough space.
#
# More information about the updates in `PopulateMergeRequestMetricsWithEventsDataImproved` class.
#
MergeRequest.all.each_batch(of: BATCH_SIZE) do |relation, index|
range = relation.pluck('MIN(id)', 'MAX(id)').first
BackgroundMigrationWorker.perform_in(index * 8.minutes, MIGRATION, range)
end
end
def down
end
end
|