debian-mirror-gitlab/app/services/ci/job_artifacts/destroy_batch_service.rb

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

201 lines
7.5 KiB
Ruby
Raw Normal View History

2021-04-29 21:17:54 +05:30
# frozen_string_literal: true
module Ci
module JobArtifacts
class DestroyBatchService
include BaseServiceUtility
include ::Gitlab::Utils::StrongMemoize
# Danger: Private - Should only be called in Ci Services that pass a batch of job artifacts
# Not for use outside of the Ci:: namespace
# Adds the passed batch of job artifacts to the `ci_deleted_objects` table
# for asyncronous destruction of the objects in Object Storage via the `Ci::DeleteObjectsService`
# and then deletes the batch of related `ci_job_artifacts` records.
# Params:
# +job_artifacts+:: A relation of job artifacts to destroy (fewer than MAX_JOB_ARTIFACT_BATCH_SIZE)
# +pick_up_at+:: When to pick up for deletion of files
# Returns:
# +Hash+:: A hash with status and destroyed_artifacts_count keys
2022-07-23 23:45:48 +05:30
def initialize(job_artifacts, pick_up_at: nil, fix_expire_at: fix_expire_at?, skip_projects_on_refresh: false)
2021-04-29 21:17:54 +05:30
@job_artifacts = job_artifacts.with_destroy_preloads.to_a
@pick_up_at = pick_up_at
2022-04-01 21:47:47 +05:30
@fix_expire_at = fix_expire_at
2022-07-23 23:45:48 +05:30
@skip_projects_on_refresh = skip_projects_on_refresh
2021-04-29 21:17:54 +05:30
end
# rubocop: disable CodeReuse/ActiveRecord
2021-06-08 01:23:25 +05:30
def execute(update_stats: true)
2022-07-23 23:45:48 +05:30
if @skip_projects_on_refresh
exclude_artifacts_undergoing_stats_refresh
else
track_artifacts_undergoing_stats_refresh
end
2022-04-01 21:47:47 +05:30
# Detect and fix artifacts that had `expire_at` wrongly backfilled by migration
# https://gitlab.com/gitlab-org/gitlab/-/merge_requests/47723
detect_and_fix_wrongly_expired_artifacts
2021-06-08 01:23:25 +05:30
return success(destroyed_artifacts_count: 0, statistics_updates: {}) if @job_artifacts.empty?
2021-04-29 21:17:54 +05:30
2021-12-11 22:18:48 +05:30
destroy_related_records(@job_artifacts)
2022-07-16 23:28:13 +05:30
destroy_around_hook(@job_artifacts) do
Ci::DeletedObject.transaction do
Ci::DeletedObject.bulk_import(@job_artifacts, @pick_up_at)
Ci::JobArtifact.id_in(@job_artifacts.map(&:id)).delete_all
end
2021-04-29 21:17:54 +05:30
end
2021-12-11 22:18:48 +05:30
after_batch_destroy_hook(@job_artifacts)
2021-04-29 21:17:54 +05:30
# This is executed outside of the transaction because it depends on Redis
2021-06-08 01:23:25 +05:30
update_project_statistics! if update_stats
2021-12-11 22:18:48 +05:30
increment_monitoring_statistics(artifacts_count, artifacts_bytes)
2021-04-29 21:17:54 +05:30
2022-08-27 11:52:29 +05:30
Gitlab::Ci::Artifacts::Logger.log_deleted(@job_artifacts, 'Ci::JobArtifacts::DestroyBatchService#execute')
2021-06-08 01:23:25 +05:30
success(destroyed_artifacts_count: artifacts_count,
2022-08-27 11:52:29 +05:30
statistics_updates: affected_project_statistics)
2021-04-29 21:17:54 +05:30
end
# rubocop: enable CodeReuse/ActiveRecord
private
2022-07-16 23:28:13 +05:30
# Overriden in EE
# :nocov:
def destroy_around_hook(artifacts)
yield
end
# :nocov:
2021-12-11 22:18:48 +05:30
# Overriden in EE
2021-04-29 21:17:54 +05:30
def destroy_related_records(artifacts); end
2021-12-11 22:18:48 +05:30
# Overriden in EE
def after_batch_destroy_hook(artifacts); end
2021-06-08 01:23:25 +05:30
# using ! here since this can't be called inside a transaction
def update_project_statistics!
affected_project_statistics.each do |project, delta|
project.increment_statistic_value(Ci::JobArtifact.project_statistics_name, delta)
end
end
def affected_project_statistics
strong_memoize(:affected_project_statistics) do
artifacts_by_project = @job_artifacts.group_by(&:project)
artifacts_by_project.each.with_object({}) do |(project, artifacts), accumulator|
delta = -artifacts.sum { |artifact| artifact.size.to_i }
accumulator[project] = delta
end
2021-04-29 21:17:54 +05:30
end
end
2021-12-11 22:18:48 +05:30
def increment_monitoring_statistics(size, bytes)
metrics.increment_destroyed_artifacts_count(size)
metrics.increment_destroyed_artifacts_bytes(bytes)
2021-04-29 21:17:54 +05:30
end
def metrics
@metrics ||= ::Gitlab::Ci::Artifacts::Metrics.new
end
def artifacts_count
strong_memoize(:artifacts_count) do
@job_artifacts.count
end
end
2021-12-11 22:18:48 +05:30
def artifacts_bytes
strong_memoize(:artifacts_bytes) do
@job_artifacts.sum { |artifact| artifact.try(:size) || 0 }
end
end
2022-04-01 21:47:47 +05:30
# This detects and fixes job artifacts that have `expire_at` wrongly backfilled by the migration
# https://gitlab.com/gitlab-org/gitlab/-/merge_requests/47723.
# These job artifacts will not be deleted and will have their `expire_at` removed.
#
# The migration would have backfilled `expire_at`
# to midnight on the 22nd of the month of the local timezone,
# storing it as UTC time in the database.
#
# If the timezone setting has changed since the migration,
# the `expire_at` stored in the database could have changed to a different local time other than midnight.
# For example:
# - changing timezone from UTC+02:00 to UTC+02:30 would change the `expire_at` in local time 00:00:00 to 00:30:00.
# - changing timezone from UTC+00:00 to UTC-01:00 would change the `expire_at` in local time 00:00:00 to 23:00:00 on the previous day (21st).
#
# Therefore job artifacts that have `expire_at` exactly on the 00, 30 or 45 minute mark
# on the dates 21, 22, 23 of the month will not be deleted.
# https://en.wikipedia.org/wiki/List_of_UTC_time_offsets
def detect_and_fix_wrongly_expired_artifacts
return unless @fix_expire_at
wrongly_expired_artifacts, @job_artifacts = @job_artifacts.partition { |artifact| wrongly_expired?(artifact) }
2022-06-21 17:19:12 +05:30
remove_expire_at(wrongly_expired_artifacts) if wrongly_expired_artifacts.any?
2022-04-01 21:47:47 +05:30
end
def fix_expire_at?
2022-07-16 23:28:13 +05:30
Feature.enabled?(:ci_detect_wrongly_expired_artifacts)
2022-04-01 21:47:47 +05:30
end
def wrongly_expired?(artifact)
return false unless artifact.expire_at.present?
2022-06-21 17:19:12 +05:30
# Although traces should never have expiration dates that don't match time & date here.
# we can explicitly exclude them by type since they should never be destroyed.
artifact.trace? || (match_date?(artifact.expire_at) && match_time?(artifact.expire_at))
2022-04-01 21:47:47 +05:30
end
def match_date?(expire_at)
[21, 22, 23].include?(expire_at.day)
end
def match_time?(expire_at)
%w[00:00.000 30:00.000 45:00.000].include?(expire_at.strftime('%M:%S.%L'))
end
def remove_expire_at(artifacts)
Ci::JobArtifact.id_in(artifacts).update_all(expire_at: nil)
Gitlab::AppLogger.info(message: "Fixed expire_at from artifacts.", fixed_artifacts_expire_at_count: artifacts.count)
end
2022-07-23 23:45:48 +05:30
def track_artifacts_undergoing_stats_refresh
project_ids = @job_artifacts.find_all do |artifact|
artifact.project.refreshing_build_artifacts_size?
end.map(&:project_id).uniq
project_ids.each do |project_id|
Gitlab::ProjectStatsRefreshConflictsLogger.warn_artifact_deletion_during_stats_refresh(
method: 'Ci::JobArtifacts::DestroyBatchService#execute',
project_id: project_id
)
end
end
def exclude_artifacts_undergoing_stats_refresh
project_ids = Set.new
@job_artifacts.reject! do |artifact|
next unless artifact.project.refreshing_build_artifacts_size?
project_ids << artifact.project_id
end
2022-08-13 15:12:31 +05:30
if project_ids.any?
Gitlab::ProjectStatsRefreshConflictsLogger.warn_skipped_artifact_deletion_during_stats_refresh(
method: 'Ci::JobArtifacts::DestroyBatchService#execute',
project_ids: project_ids
)
end
2022-07-23 23:45:48 +05:30
end
2021-04-29 21:17:54 +05:30
end
end
end
2021-06-08 01:23:25 +05:30
Ci::JobArtifacts::DestroyBatchService.prepend_mod_with('Ci::JobArtifacts::DestroyBatchService')