Merge branch 'eb-recalculate-artifacts-size-script' into 'master'

Add script to recalculate project statistics build artifacts size

See merge request gitlab-org/gitlab!81306

app/models/concerns/counter_attribute.rb

@@ -102,9 +102,7 @@ module CounterAttribute
 
     run_after_commit_or_now do
       if counter_attribute_enabled?(attribute)
-        redis_state do |redis|
-          redis.incrby(counter_key(attribute), increment)
-        end
+        increment_counter(attribute, increment)
 
         FlushCounterIncrementsWorker.perform_in(WORKER_DELAY, self.class.name, self.id, attribute)
       else
@@ -115,6 +113,28 @@ module CounterAttribute
     true
   end
 
+  def increment_counter(attribute, increment)
+    if counter_attribute_enabled?(attribute)
+      redis_state do |redis|
+        redis.incrby(counter_key(attribute), increment)
+      end
+    end
+  end
+
+  def clear_counter!(attribute)
+    if counter_attribute_enabled?(attribute)
+      redis_state { |redis| redis.del(counter_key(attribute)) }
+    end
+  end
+
+  def get_counter_value(attribute)
+    if counter_attribute_enabled?(attribute)
+      redis_state do |redis|
+        redis.get(counter_key(attribute)).to_i
+      end
+    end
+  end
+
   def counter_key(attribute)
     "project:{#{project_id}}:counters:#{self.class}:#{id}:#{attribute}"
   end

app/models/projects/build_artifacts_size_refresh.rb

+# frozen_string_literal: true
+
+module Projects
+  class BuildArtifactsSizeRefresh < ApplicationRecord
+    include BulkInsertSafe
+
+    STALE_WINDOW = 3.days
+
+    self.table_name = 'project_build_artifacts_size_refreshes'
+
+    belongs_to :project
+
+    validates :project, presence: true
+
+    STATES = {
+      created: 1,
+      running: 2,
+      pending: 3
+    }.freeze
+
+    state_machine :state, initial: :created do
+      # created -> running <-> pending
+      state :created, value: STATES[:created]
+      state :running, value: STATES[:running]
+      state :pending, value: STATES[:pending]
+
+      event :process do
+        transition [:created, :pending, :running] => :running
+      end
+
+      event :requeue do
+        transition running: :pending
+      end
+
+      # set it only the first time we execute the refresh
+      before_transition created: :running do |refresh|
+        refresh.reset_project_statistics!
+        refresh.refresh_started_at = Time.zone.now
+      end
+
+      before_transition running: any do |refresh, transition|
+        refresh.updated_at = Time.zone.now
+      end
+
+      before_transition running: :pending do |refresh, transition|
+        refresh.last_job_artifact_id = transition.args.first
+      end
+    end
+
+    scope :stale, -> { with_state(:running).where('updated_at < ?', STALE_WINDOW.ago) }
+    scope :remaining, -> { with_state(:created, :pending).or(stale) }
+
+    def self.enqueue_refresh(projects)
+      now = Time.zone.now
+
+      records = Array(projects).map do |project|
+        new(project: project, state: STATES[:created], created_at: now, updated_at: now)
+      end
+
+      bulk_insert!(records, skip_duplicates: true)
+    end
+
+    def self.process_next_refresh!
+      next_refresh = nil
+
+      transaction do
+        next_refresh = remaining
+          .order(:state, :updated_at)
+          .lock('FOR UPDATE SKIP LOCKED')
+          .take
+
+        next_refresh&.process!
+      end
+
+      next_refresh
+    end
+
+    def reset_project_statistics!
+      statistics = project.statistics
+      statistics.update!(build_artifacts_size: 0)
+      statistics.clear_counter!(:build_artifacts_size)
+    end
+
+    def next_batch(limit:)
+      project.job_artifacts.select(:id, :size)
+        .where('created_at <= ? AND id > ?', refresh_started_at, last_job_artifact_id.to_i)
+        .order(:created_at)
+        .limit(limit)
+    end
+  end
+end

app/services/projects/refresh_build_artifacts_size_statistics_service.rb

+# frozen_string_literal: true
+
+module Projects
+  class RefreshBuildArtifactsSizeStatisticsService
+    BATCH_SIZE = 1000
+
+    def execute
+      refresh = Projects::BuildArtifactsSizeRefresh.process_next_refresh!
+      return unless refresh
+
+      batch = refresh.next_batch(limit: BATCH_SIZE).to_a
+
+      if batch.any?
+        # We are doing the sum in ruby because the query takes too long when done in SQL
+        total_artifacts_size = batch.sum(&:size)
+
+        Projects::BuildArtifactsSizeRefresh.transaction do
+          # Mark the refresh ready for another worker to pick up and process the next batch
+          refresh.requeue!(batch.last.id)
+
+          refresh.project.statistics.delayed_increment_counter(:build_artifacts_size, total_artifacts_size)
+        end
+      else
+        # Remove the refresh job from the table if there are no more
+        # remaining job artifacts to calculate for the given project.
+        refresh.destroy!
+      end
+
+      refresh
+    end
+  end
+end

app/workers/all_queues.yml

@@ -561,6 +561,15 @@
   :weight: 1
   :idempotent:
   :tags: []
+- :name: cronjob:projects_schedule_refresh_build_artifacts_size_statistics
+  :worker_name: Projects::ScheduleRefreshBuildArtifactsSizeStatisticsWorker
+  :feature_category: :build_artifacts
+  :has_external_dependencies:
+  :urgency: :low
+  :resource_boundary: :unknown
+  :weight: 1
+  :idempotent: true
+  :tags: []
 - :name: cronjob:prune_old_events
   :worker_name: PruneOldEventsWorker
   :feature_category: :users
@@ -2803,6 +2812,15 @@
   :weight: 1
   :idempotent: true
   :tags: []
+- :name: projects_refresh_build_artifacts_size_statistics
+  :worker_name: Projects::RefreshBuildArtifactsSizeStatisticsWorker
+  :feature_category: :build_artifacts
+  :has_external_dependencies:
+  :urgency: :low
+  :resource_boundary: :unknown
+  :weight: 1
+  :idempotent: true
+  :tags: []
 - :name: projects_schedule_bulk_repository_shard_moves
   :worker_name: Projects::ScheduleBulkRepositoryShardMovesWorker
   :feature_category: :gitaly

app/workers/projects/refresh_build_artifacts_size_statistics_worker.rb

+# frozen_string_literal: true
+
+module Projects
+  class RefreshBuildArtifactsSizeStatisticsWorker
+    include ApplicationWorker
+    include LimitedCapacity::Worker
+
+    MAX_RUNNING_LOW = 2
+    MAX_RUNNING_MEDIUM = 20
+    MAX_RUNNING_HIGH = 50
+
+    data_consistency :always
+
+    feature_category :build_artifacts
+
+    idempotent!
+
+    def perform_work(*args)
+      refresh = Projects::RefreshBuildArtifactsSizeStatisticsService.new.execute
+      return unless refresh
+
+      log_extra_metadata_on_done(:project_id, refresh.project_id)
+      log_extra_metadata_on_done(:last_job_artifact_id, refresh.last_job_artifact_id)
+      log_extra_metadata_on_done(:last_batch, refresh.destroyed?)
+      log_extra_metadata_on_done(:refresh_started_at, refresh.refresh_started_at)
+    end
+
+    def remaining_work_count(*args)
+      # LimitedCapacity::Worker only needs to know if there is work left to do
+      # so we can get by with an EXISTS query rather than a count.
+      # https://gitlab.com/gitlab-org/gitlab/-/issues/356167
+      if Projects::BuildArtifactsSizeRefresh.remaining.any?
+        1
+      else
+        0
+      end
+    end
+
+    def max_running_jobs
+      if ::Feature.enabled?(:projects_build_artifacts_size_refresh_high)
+        MAX_RUNNING_HIGH
+      elsif ::Feature.enabled?(:projects_build_artifacts_size_refresh_medium)
+        MAX_RUNNING_MEDIUM
+      elsif ::Feature.enabled?(:projects_build_artifacts_size_refresh_low)
+        MAX_RUNNING_LOW
+      else
+        0
+      end
+    end
+  end
+end

app/workers/projects/schedule_refresh_build_artifacts_size_statistics_worker.rb

+# frozen_string_literal: true
+
+module Projects
+  class ScheduleRefreshBuildArtifactsSizeStatisticsWorker
+    include ApplicationWorker
+    include CronjobQueue # rubocop:disable Scalability/CronWorkerContext
+
+    data_consistency :always
+
+    feature_category :build_artifacts
+
+    idempotent!
+
+    def perform
+      Projects::RefreshBuildArtifactsSizeStatisticsWorker.perform_with_capacity
+    end
+  end
+end

config/feature_flags/development/projects_build_artifacts_size_refresh_high.yml

+---
+name: projects_build_artifacts_size_refresh_high
+introduced_by_url: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/81306
+rollout_issue_url: https://gitlab.com/gitlab-org/gitlab/-/issues/356018
+milestone: '14.9'
+type: development
+group: group::pipeline insights
+default_enabled: false

config/feature_flags/development/projects_build_artifacts_size_refresh_low.yml

+---
+name: projects_build_artifacts_size_refresh_low
+introduced_by_url: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/81306
+rollout_issue_url: https://gitlab.com/gitlab-org/gitlab/-/issues/356018
+milestone: '14.9'
+type: development
+group: group::pipeline insights
+default_enabled: false

config/feature_flags/development/projects_build_artifacts_size_refresh_medium.yml

+---
+name: projects_build_artifacts_size_refresh_medium
+introduced_by_url: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/81306
+rollout_issue_url: https://gitlab.com/gitlab-org/gitlab/-/issues/356018
+milestone: '14.9'
+type: development
+group: group::pipeline insights
+default_enabled: false

config/initializers/1_settings.rb

@@ -620,6 +620,9 @@ Settings.cron_jobs['issues_reschedule_stuck_issue_rebalances']['job_class'] = 'I
 Settings.cron_jobs['clusters_integrations_check_prometheus_health_worker'] ||= Settingslogic.new({})
 Settings.cron_jobs['clusters_integrations_check_prometheus_health_worker']['cron'] ||= '0 * * * *'
 Settings.cron_jobs['clusters_integrations_check_prometheus_health_worker']['job_class'] = 'Clusters::Integrations::CheckPrometheusHealthWorker'
+Settings.cron_jobs['projects_schedule_refresh_build_artifacts_size_statistics_worker'] ||= Settingslogic.new({})
+Settings.cron_jobs['projects_schedule_refresh_build_artifacts_size_statistics_worker']['cron'] ||= '2/17 * * * *'
+Settings.cron_jobs['projects_schedule_refresh_build_artifacts_size_statistics_worker']['job_class'] = 'Projects::ScheduleRefreshBuildArtifactsSizeStatisticsWorker'
 
 Gitlab.ee do
   Settings.cron_jobs['analytics_devops_adoption_create_all_snapshots_worker'] ||= Settingslogic.new({})

config/sidekiq_queues.yml

@@ -363,6 +363,8 @@
   - 1
 - - projects_process_sync_events
   - 1
+- - projects_refresh_build_artifacts_size_statistics
+  - 1
 - - projects_schedule_bulk_repository_shard_moves
   - 1
 - - projects_update_repository_storage

db/migrate/20220301175426_create_project_build_artifacts_size_refresh.rb

+# frozen_string_literal: true
+
+class CreateProjectBuildArtifactsSizeRefresh < Gitlab::Database::Migration[1.0]
+  enable_lock_retries!
+
+  CREATED_STATE = 1
+
+  def change
+    create_table :project_build_artifacts_size_refreshes do |t|
+      t.references :project, index: { unique: true }, foreign_key: { on_delete: :cascade }, null: false
+      t.bigint :last_job_artifact_id, null: true
+      t.integer :state, null: false, default: CREATED_STATE, limit: 1
+      t.datetime_with_timezone :refresh_started_at, null: true
+      t.timestamps_with_timezone null: false
+
+      # We will use this index for 2 purposes:
+      # - for finding rows with state = :waiting
+      # - for finding rows with state = :running and updated_at < x.days.ago
+      #   which we can use to find jobs that were not able to complete and considered
+      #   stale so we can retry
+      t.index [:state, :updated_at], name: 'idx_build_artifacts_size_refreshes_state_updated_at'
+    end
+  end
+end

db/post_migrate/20220314162342_add_index_ci_job_artifacts_project_id_created_at.rb

+# frozen_string_literal: true
+
+class AddIndexCiJobArtifactsProjectIdCreatedAt < Gitlab::Database::Migration[1.0]
+  INDEX_NAME = 'index_ci_job_artifacts_on_id_project_id_and_created_at'
+
+  disable_ddl_transaction!
+
+  def up
+    add_concurrent_index :ci_job_artifacts, [:project_id, :created_at, :id], name: INDEX_NAME
+  end
+
+  def down
+    remove_concurrent_index_by_name :ci_job_artifacts, INDEX_NAME
+  end
+end

db/schema_migrations/20220301175426

+d0a8daf9fb9892fc92b03f13de4d7e470e5c54f03b09f887cdd45bc5eb9a7e37
\ No newline at end of file

db/schema_migrations/20220314162342

+7992448797888fd69d1e5cd4f2602e5a2b49a57052c50b19522f37d711c9f2f2
\ No newline at end of file

db/structure.sql

@@ -18974,6 +18974,25 @@ CREATE SEQUENCE project_auto_devops_id_seq
 
 ALTER SEQUENCE project_auto_devops_id_seq OWNED BY project_auto_devops.id;
 
+CREATE TABLE project_build_artifacts_size_refreshes (
+    id bigint NOT NULL,
+    project_id bigint NOT NULL,
+    last_job_artifact_id bigint,
+    state smallint DEFAULT 1 NOT NULL,
+    refresh_started_at timestamp with time zone,
+    created_at timestamp with time zone NOT NULL,
+    updated_at timestamp with time zone NOT NULL
+);
+
+CREATE SEQUENCE project_build_artifacts_size_refreshes_id_seq
+    START WITH 1
+    INCREMENT BY 1
+    NO MINVALUE
+    NO MAXVALUE
+    CACHE 1;
+
+ALTER SEQUENCE project_build_artifacts_size_refreshes_id_seq OWNED BY project_build_artifacts_size_refreshes.id;
+
 CREATE TABLE project_ci_cd_settings (
     id integer NOT NULL,
     project_id integer NOT NULL,
@@ -22870,6 +22889,8 @@ ALTER TABLE ONLY project_aliases ALTER COLUMN id SET DEFAULT nextval('project_al
 
 ALTER TABLE ONLY project_auto_devops ALTER COLUMN id SET DEFAULT nextval('project_auto_devops_id_seq'::regclass);
 
+ALTER TABLE ONLY project_build_artifacts_size_refreshes ALTER COLUMN id SET DEFAULT nextval('project_build_artifacts_size_refreshes_id_seq'::regclass);
+
 ALTER TABLE ONLY project_ci_cd_settings ALTER COLUMN id SET DEFAULT nextval('project_ci_cd_settings_id_seq'::regclass);
 
 ALTER TABLE ONLY project_ci_feature_usages ALTER COLUMN id SET DEFAULT nextval('project_ci_feature_usages_id_seq'::regclass);
@@ -24940,6 +24961,9 @@ ALTER TABLE ONLY project_authorizations
 ALTER TABLE ONLY project_auto_devops
     ADD CONSTRAINT project_auto_devops_pkey PRIMARY KEY (id);
 
+ALTER TABLE ONLY project_build_artifacts_size_refreshes
+    ADD CONSTRAINT project_build_artifacts_size_refreshes_pkey PRIMARY KEY (id);
+
 ALTER TABLE ONLY project_ci_cd_settings
     ADD CONSTRAINT project_ci_cd_settings_pkey PRIMARY KEY (id);
 
@@ -26398,6 +26422,8 @@ CREATE INDEX idx_audit_events_part_on_entity_id_desc_author_id_created_at ON ONL
 
 CREATE INDEX idx_award_emoji_on_user_emoji_name_awardable_type_awardable_id ON award_emoji USING btree (user_id, name, awardable_type, awardable_id);
 
+CREATE INDEX idx_build_artifacts_size_refreshes_state_updated_at ON project_build_artifacts_size_refreshes USING btree (state, updated_at);
+
 CREATE INDEX idx_ci_pipelines_artifacts_locked ON ci_pipelines USING btree (ci_ref_id, id) WHERE (locked = 1);
 
 CREATE INDEX idx_container_exp_policies_on_project_id_next_run_at ON container_expiration_policies USING btree (project_id, next_run_at) WHERE (enabled = true);
@@ -26920,6 +26946,8 @@ CREATE INDEX index_ci_job_artifacts_on_file_store ON ci_job_artifacts USING btre
 
 CREATE INDEX index_ci_job_artifacts_on_file_type_for_devops_adoption ON ci_job_artifacts USING btree (file_type, project_id, created_at) WHERE (file_type = ANY (ARRAY[5, 6, 8, 23]));
 
+CREATE INDEX index_ci_job_artifacts_on_id_project_id_and_created_at ON ci_job_artifacts USING btree (project_id, created_at, id);
+
 CREATE INDEX index_ci_job_artifacts_on_id_project_id_and_file_type ON ci_job_artifacts USING btree (project_id, file_type, id);
 
 CREATE UNIQUE INDEX index_ci_job_artifacts_on_job_id_and_file_type ON ci_job_artifacts USING btree (job_id, file_type);
@@ -28528,6 +28556,8 @@ CREATE INDEX index_project_aliases_on_project_id ON project_aliases USING btree
 
 CREATE UNIQUE INDEX index_project_auto_devops_on_project_id ON project_auto_devops USING btree (project_id);
 
+CREATE UNIQUE INDEX index_project_build_artifacts_size_refreshes_on_project_id ON project_build_artifacts_size_refreshes USING btree (project_id);
+
 CREATE UNIQUE INDEX index_project_ci_cd_settings_on_project_id ON project_ci_cd_settings USING btree (project_id);
 
 CREATE UNIQUE INDEX index_project_ci_feature_usages_unique_columns ON project_ci_feature_usages USING btree (project_id, feature, default_branch);
@@ -32667,6 +32697,9 @@ ALTER TABLE ONLY list_user_preferences
 ALTER TABLE ONLY merge_request_cleanup_schedules
     ADD CONSTRAINT fk_rails_92dd0e705c FOREIGN KEY (merge_request_id) REFERENCES merge_requests(id) ON DELETE CASCADE;
 
+ALTER TABLE ONLY project_build_artifacts_size_refreshes
+    ADD CONSTRAINT fk_rails_936db5fc44 FOREIGN KEY (project_id) REFERENCES projects(id) ON DELETE CASCADE;
+
 ALTER TABLE ONLY board_labels
     ADD CONSTRAINT fk_rails_9374a16edd FOREIGN KEY (board_id) REFERENCES boards(id) ON DELETE CASCADE;
 

lib/gitlab/database/gitlab_schemas.yml

@@ -401,6 +401,7 @@ project_alerting_settings: :gitlab_main
 project_aliases: :gitlab_main
 project_authorizations: :gitlab_main
 project_auto_devops: :gitlab_main
+project_build_artifacts_size_refreshes: :gitlab_main
 project_ci_cd_settings: :gitlab_main
 project_ci_feature_usages: :gitlab_main
 project_compliance_framework_settings: :gitlab_main

lib/tasks/gitlab/refresh_project_statistics_build_artifacts_size.rake

+# frozen_string_literal: true
+
+namespace :gitlab do
+  desc "GitLab | Refresh build artifacts size project statistics for given project IDs"
+
+  BUILD_ARTIFACTS_SIZE_REFRESH_ENQUEUE_BATCH_SIZE = 500
+
+  task :refresh_project_statistics_build_artifacts_size, [:project_ids] => :environment do |_t, args|
+    project_ids = []
+    project_ids = $stdin.read.split unless $stdin.tty?
+    project_ids = args.project_ids.to_s.split unless project_ids.any?
+
+    if project_ids.any?
+      project_ids.in_groups_of(BUILD_ARTIFACTS_SIZE_REFRESH_ENQUEUE_BATCH_SIZE) do |ids|
+        projects = Project.where(id: ids)
+        Projects::BuildArtifactsSizeRefresh.enqueue_refresh(projects)
+      end
+      puts 'Done.'.green
+    else
+      puts 'Please provide a string of space-separated project IDs as the argument or through the STDIN'.red
+    end
+  end
+end

spec/db/schema_spec.rb

@@ -67,6 +67,7 @@ RSpec.describe 'Database schema' do
     oauth_access_tokens: %w[resource_owner_id application_id],
     oauth_applications: %w[owner_id],
     product_analytics_events_experimental: %w[event_id txn_id user_id],
+    project_build_artifacts_size_refreshes: %w[last_job_artifact_id],
     project_group_links: %w[group_id],
     project_statistics: %w[namespace_id],
     projects: %w[creator_id ci_id mirror_user_id],

spec/factories/projects/build_artifacts_size_refreshes.rb

+# frozen_string_literal: true
+
+FactoryBot.define do
+  factory :project_build_artifacts_size_refresh, class: 'Projects::BuildArtifactsSizeRefresh' do
+    project factory: :project
+
+    trait :created do
+      state { Projects::BuildArtifactsSizeRefresh::STATES[:created] }
+    end
+
+    trait :pending do
+      state { Projects::BuildArtifactsSizeRefresh::STATES[:pending] }
+      refresh_started_at { Time.zone.now }
+    end
+
+    trait :running do
+      state { Projects::BuildArtifactsSizeRefresh::STATES[:running] }
+      refresh_started_at { Time.zone.now }
+    end
+
+    trait :stale do
+      running
+      refresh_started_at { 30.days.ago }
+      updated_at { 30.days.ago }
+    end
+  end
+end

spec/models/projects/build_artifacts_size_refresh_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+RSpec.describe Projects::BuildArtifactsSizeRefresh, type: :model do
+  describe 'associations' do
+    it { is_expected.to belong_to(:project) }
+  end
+
+  it_behaves_like 'having unique enum values'
+
+  describe 'validations' do
+    it { is_expected.to validate_presence_of(:project) }
+  end
+
+  describe 'scopes' do
+    let_it_be(:refresh_1) { create(:project_build_artifacts_size_refresh, :running, updated_at: 4.days.ago) }
+    let_it_be(:refresh_2) { create(:project_build_artifacts_size_refresh, :running, updated_at: 2.days.ago) }
+    let_it_be(:refresh_3) { create(:project_build_artifacts_size_refresh, :pending) }
+    let_it_be(:refresh_4) { create(:project_build_artifacts_size_refresh, :created) }
+
+    describe 'stale' do
+      it 'returns records in running state and has not been updated for more than 3 days' do
+        expect(described_class.stale).to eq([refresh_1])
+      end
+    end
+
+    describe 'remaining' do
+      it 'returns stale, created, and pending records' do
+        expect(described_class.remaining).to match_array([refresh_1, refresh_3, refresh_4])
+      end
+    end
+  end
+
+  describe 'state machine', :clean_gitlab_redis_shared_state do
+    around do |example|
+      freeze_time { example.run }
+    end
+
+    let(:now) { Time.zone.now }
+
+    describe 'initial state' do
+      let(:refresh) { create(:project_build_artifacts_size_refresh) }
+
+      it 'defaults to created' do
+        expect(refresh).to be_created
+      end
+    end
+
+    describe '#process!' do
+      context 'when refresh state is created' do
+        let!(:refresh) do
+          create(
+            :project_build_artifacts_size_refresh,
+            :created,
+            updated_at: 2.days.ago,
+            refresh_started_at: nil,
+            last_job_artifact_id: nil
+          )
+        end
+
+        before do
+          stats = create(:project_statistics, project: refresh.project, build_artifacts_size: 120)
+          stats.increment_counter(:build_artifacts_size, 30)
+        end
+
+        it 'transitions the state to running' do
+          expect { refresh.process! }.to change { refresh.reload.state }.to(described_class::STATES[:running])
+        end
+
+        it 'sets the refresh_started_at' do
+          expect { refresh.process! }.to change { refresh.reload.refresh_started_at.to_i }.to(now.to_i)
+        end
+
+        it 'bumps the updated_at' do
+          expect { refresh.process! }.to change { refresh.reload.updated_at.to_i }.to(now.to_i)
+        end
+
+        it 'resets the build artifacts size stats' do
+          expect { refresh.process! }.to change { refresh.project.statistics.reload.build_artifacts_size }.to(0)
+        end
+
+        it 'resets the counter attribute to zero' do
+          expect { refresh.process! }.to change { refresh.project.statistics.get_counter_value(:build_artifacts_size) }.to(0)
+        end
+      end
+
+      context 'when refresh state is pending' do
+        let!(:refresh) do
+          create(
+            :project_build_artifacts_size_refresh,
+            :pending,
+            updated_at: 2.days.ago
+          )
+        end
+
+        before do
+          create(:project_statistics, project: refresh.project)
+        end
+
+        it 'transitions the state to running' do
+          expect { refresh.process! }.to change { refresh.reload.state }.to(described_class::STATES[:running])
+        end
+
+        it 'bumps the updated_at' do
+          expect { refresh.process! }.to change { refresh.reload.updated_at.to_i }.to(now.to_i)
+        end
+      end
+
+      context 'when refresh state is running' do
+        let!(:refresh) do
+          create(
+            :project_build_artifacts_size_refresh,
+            :running,
+            updated_at: 2.days.ago
+          )
+        end
+
+        before do
+          create(:project_statistics, project: refresh.project)
+        end
+
+        it 'keeps the state at running' do
+          expect { refresh.process! }.not_to change { refresh.reload.state }
+        end
+
+        it 'bumps the updated_at' do
+          # If this was a stale job, we want to bump the updated at now so that
+          # it won't be picked up by another worker while we're recalculating
+          expect { refresh.process! }.to change { refresh.reload.updated_at.to_i }.to(now.to_i)
+        end
+      end
+    end
+
+    describe '#requeue!' do
+      let!(:refresh) do
+        create(
+          :project_build_artifacts_size_refresh,
+          :running,
+          updated_at: 2.days.ago,
+          last_job_artifact_id: 111
+        )
+      end
+
+      let(:last_job_artifact_id) { 123 }
+
+      it 'transitions refresh state from running to pending' do
+        expect { refresh.requeue!(last_job_artifact_id) }.to change { refresh.reload.state }.to(described_class::STATES[:pending])
+      end
+
+      it 'bumps updated_at' do
+        expect { refresh.requeue!(last_job_artifact_id) }.to change { refresh.reload.updated_at.to_i }.to(now.to_i)
+      end
+
+      it 'updates last_job_artifact_id' do
+        expect { refresh.requeue!(last_job_artifact_id) }.to change { refresh.reload.last_job_artifact_id.to_i }.to(last_job_artifact_id)
+      end
+    end
+  end
+
+  describe '.process_next_refresh!' do
+    let!(:refresh_running) { create(:project_build_artifacts_size_refresh, :running) }
+    let!(:refresh_created) { create(:project_build_artifacts_size_refresh, :created) }
+    let!(:refresh_stale) { create(:project_build_artifacts_size_refresh, :stale) }
+    let!(:refresh_pending) { create(:project_build_artifacts_size_refresh, :pending) }
+
+    subject(:processed_refresh) { described_class.process_next_refresh! }
+
+    it 'picks the first record from the remaining work' do
+      expect(processed_refresh).to eq(refresh_created)
+      expect(processed_refresh.reload).to be_running
+    end
+  end
+
+  describe '.enqueue_refresh' do
+    let_it_be(:project_1) { create(:project) }
+    let_it_be(:project_2) { create(:project) }
+
+    let(:projects) { [project_1, project_1, project_2] }
+
+    it 'creates refresh records for each given project, skipping duplicates' do
+      expect { described_class.enqueue_refresh(projects) }
+        .to change { described_class.count }.from(0).to(2)
+
+      expect(described_class.first).to have_attributes(
+        project_id: project_1.id,
+        last_job_artifact_id: nil,
+        refresh_started_at: nil,
+        state: described_class::STATES[:created]
+      )
+
+      expect(described_class.last).to have_attributes(
+        project_id: project_2.id,
+        last_job_artifact_id: nil,
+        refresh_started_at: nil,
+        state: described_class::STATES[:created]
+      )
+    end
+  end
+
+  describe '#next_batch' do
+    let!(:project) { create(:project) }
+    let!(:artifact_1) { create(:ci_job_artifact, project: project, created_at: 14.days.ago) }
+    let!(:artifact_2) { create(:ci_job_artifact, project: project, created_at: 13.days.ago) }
+    let!(:artifact_3) { create(:ci_job_artifact, project: project, created_at: 12.days.ago) }
+
+    # This should not be included in the recalculation as it is created later than the refresh start time
+    let!(:future_artifact) { create(:ci_job_artifact, project: project, size: 8, created_at: refresh.refresh_started_at + 1.second) }
+
+    let!(:refresh) do
+      create(
+        :project_build_artifacts_size_refresh,
+        :pending,
+        project: project,
+        updated_at: 2.days.ago,
+        refresh_started_at: 10.days.ago,
+        last_job_artifact_id: artifact_1.id
+      )
+    end
+
+    subject(:batch) { refresh.next_batch(limit: 3) }
+
+    it 'returns the job artifact records that were created not later than the refresh_started_at and IDs greater than the last_job_artifact_id' do
+      expect(batch).to eq([artifact_2, artifact_3])
+    end
+  end
+end

spec/services/projects/refresh_build_artifacts_size_statistics_service_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+RSpec.describe Projects::RefreshBuildArtifactsSizeStatisticsService, :clean_gitlab_redis_shared_state do
+  let(:service) { described_class.new }
+
+  describe '#execute' do
+    let_it_be(:project) { create(:project) }
+
+    let_it_be(:artifact_1) { create(:ci_job_artifact, project: project, size: 1, created_at: 14.days.ago) }
+    let_it_be(:artifact_2) { create(:ci_job_artifact, project: project, size: 2, created_at: 13.days.ago) }
+    let_it_be(:artifact_3) { create(:ci_job_artifact, project: project, size: 5, created_at: 12.days.ago) }
+
+    # This should not be included in the recalculation as it is created later than the refresh start time
+    let_it_be(:future_artifact) { create(:ci_job_artifact, project: project, size: 8, created_at: 2.days.from_now) }
+
+    let!(:refresh) do
+      create(
+        :project_build_artifacts_size_refresh,
+        :created,
+        project: project,
+        updated_at: 2.days.ago,
+        refresh_started_at: nil,
+        last_job_artifact_id: nil
+      )
+    end
+
+    let(:now) { Time.zone.now }
+
+    around do |example|
+      freeze_time { example.run }
+    end
+
+    before do
+      stub_const("#{described_class}::BATCH_SIZE", 2)
+
+      stats = create(:project_statistics, project: project, build_artifacts_size: 120)
+      stats.increment_counter(:build_artifacts_size, 30)
+    end
+
+    it 'resets the build artifacts size stats' do
+      expect { service.execute }.to change { project.statistics.reload.build_artifacts_size }.to(0)
+    end
+
+    it 'increments the counter attribute by the total size of the current batch of artifacts' do
+      expect { service.execute }.to change { project.statistics.get_counter_value(:build_artifacts_size) }.to(3)
+    end
+
+    it 'updates the last_job_artifact_id to the ID of the last artifact from the batch' do
+      expect { service.execute }.to change { refresh.reload.last_job_artifact_id.to_i }.to(artifact_2.id)
+    end
+
+    it 'requeues the refresh job' do
+      service.execute
+      expect(refresh.reload).to be_pending
+    end
+
+    context 'when an error happens after the recalculation has started' do
+      let!(:refresh) do
+        create(
+          :project_build_artifacts_size_refresh,
+          :pending,
+          project: project,
+          last_job_artifact_id: artifact_2.id
+        )
+      end
+
+      before do
+        allow(Gitlab::Redis::SharedState).to receive(:with).and_raise(StandardError, 'error')
+
+        expect { service.execute }.to raise_error(StandardError)
+      end
+
+      it 'keeps the last_job_artifact_id unchanged' do
+        expect(refresh.reload.last_job_artifact_id).to eq(artifact_2.id)
+      end
+
+      it 'keeps the state of the refresh record at running' do
+        expect(refresh.reload).to be_running
+      end
+    end
+
+    context 'when there are no more artifacts to recalculate for the next refresh job' do
+      let!(:refresh) do
+        create(
+          :project_build_artifacts_size_refresh,
+          :pending,
+          project: project,
+          updated_at: 2.days.ago,
+          refresh_started_at: now,
+          last_job_artifact_id: artifact_3.id
+        )
+      end
+
+      it 'deletes the refresh record' do
+        service.execute
+        expect(Projects::BuildArtifactsSizeRefresh.where(id: refresh.id)).not_to exist
+      end
+    end
+  end
+end

spec/tasks/gitlab/refresh_project_statistics_build_artifacts_size_rake_spec.rb

+# frozen_string_literal: true
+
+require 'rake_helper'
+
+RSpec.describe 'gitlab:refresh_project_statistics_build_artifacts_size rake task', :silence_stdout do
+  let(:rake_task) { 'gitlab:refresh_project_statistics_build_artifacts_size' }
+
+  describe 'enqueuing build artifacts size statistics refresh for given list of project IDs' do
+    let_it_be(:project_1) { create(:project) }
+    let_it_be(:project_2) { create(:project) }
+    let_it_be(:project_3) { create(:project) }
+
+    let(:string_of_ids) { "#{project_1.id} #{project_2.id} #{project_3.id} 999999" }
+
+    before do
+      Rake.application.rake_require('tasks/gitlab/refresh_project_statistics_build_artifacts_size')
+
+      stub_const("BUILD_ARTIFACTS_SIZE_REFRESH_ENQUEUE_BATCH_SIZE", 2)
+    end
+
+    context 'when given a list of space-separated IDs through STDIN' do
+      before do
+        allow($stdin).to receive(:tty?).and_return(false)
+        allow($stdin).to receive(:read).and_return(string_of_ids)
+      end
+
+      it 'enqueues the projects for refresh' do
+        expect { run_rake_task(rake_task) }.to output(/Done/).to_stdout
+
+        expect(Projects::BuildArtifactsSizeRefresh.all.map(&:project)).to match_array([project_1, project_2, project_3])
+      end
+    end
+
+    context 'when given a list of space-separated IDs through rake argument' do
+      it 'enqueues the projects for refresh' do
+        expect { run_rake_task(rake_task, string_of_ids) }.to output(/Done/).to_stdout
+
+        expect(Projects::BuildArtifactsSizeRefresh.all.map(&:project)).to match_array([project_1, project_2, project_3])
+      end
+    end
+
+    context 'when not given any IDs' do
+      it 'returns an error message' do
+        expect { run_rake_task(rake_task) }.to output(/Please provide a string of space-separated project IDs/).to_stdout
+      end
+    end
+  end
+end

spec/workers/every_sidekiq_worker_spec.rb

@@ -395,6 +395,7 @@ RSpec.describe 'Every Sidekiq worker' do
         'Projects::PostCreationWorker' => 3,
         'Projects::ScheduleBulkRepositoryShardMovesWorker' => 3,
         'Projects::UpdateRepositoryStorageWorker' => 3,
+        'Projects::RefreshBuildArtifactsSizeStatisticsWorker' => 0,
         'Prometheus::CreateDefaultAlertsWorker' => 3,
         'PropagateIntegrationGroupWorker' => 3,
         'PropagateIntegrationInheritDescendantWorker' => 3,

spec/workers/projects/refresh_build_artifacts_size_statistics_worker_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+RSpec.describe Projects::RefreshBuildArtifactsSizeStatisticsWorker do
+  let(:worker) { described_class.new }
+
+  describe '#perform_work' do
+    before do
+      expect_next_instance_of(Projects::RefreshBuildArtifactsSizeStatisticsService) do |instance|
+        expect(instance).to receive(:execute).and_return(refresh)
+      end
+    end
+
+    context 'when refresh job is present' do
+      let(:refresh) do
+        build(
+          :project_build_artifacts_size_refresh,
+          :running,
+          project_id: 77,
+          last_job_artifact_id: 123
+        )
+      end
+
+      it 'logs refresh information' do
+        expect(worker).to receive(:log_extra_metadata_on_done).with(:project_id, refresh.project_id)
+        expect(worker).to receive(:log_extra_metadata_on_done).with(:last_job_artifact_id, refresh.last_job_artifact_id)
+        expect(worker).to receive(:log_extra_metadata_on_done).with(:last_batch, refresh.destroyed?)
+        expect(worker).to receive(:log_extra_metadata_on_done).with(:refresh_started_at, refresh.refresh_started_at)
+
+        worker.perform_work
+      end
+    end
+
+    context 'when refresh job is not present' do
+      let(:refresh) { nil }
+
+      it 'logs refresh information' do
+        expect(worker).not_to receive(:log_extra_metadata_on_done)
+
+        worker.perform_work
+      end
+    end
+  end
+
+  describe '#remaining_work_count' do
+    subject { worker.remaining_work_count }
+
+    context 'and there are remaining refresh jobs' do
+      before do
+        create_list(:project_build_artifacts_size_refresh, 2, :pending)
+      end
+
+      it { is_expected.to eq(1) }
+    end
+
+    context 'and there are no remaining refresh jobs' do
+      it { is_expected.to eq(0) }
+    end
+  end
+
+  describe '#max_running_jobs' do
+    subject { worker.max_running_jobs }
+
+    context 'when all projects_build_artifacts_size_refresh flags are enabled' do
+      it { is_expected.to eq(described_class::MAX_RUNNING_HIGH) }
+    end
+
+    context 'when projects_build_artifacts_size_refresh_high flags is disabled' do
+      before do
+        stub_feature_flags(projects_build_artifacts_size_refresh_high: false)
+      end
+
+      it { is_expected.to eq(described_class::MAX_RUNNING_MEDIUM) }
+    end
+
+    context 'when projects_build_artifacts_size_refresh_high and projects_build_artifacts_size_refresh_medium flags are disabled' do
+      before do
+        stub_feature_flags(projects_build_artifacts_size_refresh_high: false)
+        stub_feature_flags(projects_build_artifacts_size_refresh_medium: false)
+      end
+
+      it { is_expected.to eq(described_class::MAX_RUNNING_LOW) }
+    end
+
+    context 'when all projects_build_artifacts_size_refresh flags are disabled' do
+      before do
+        stub_feature_flags(projects_build_artifacts_size_refresh_low: false)
+        stub_feature_flags(projects_build_artifacts_size_refresh_medium: false)
+        stub_feature_flags(projects_build_artifacts_size_refresh_high: false)
+      end
+
+      it { is_expected.to eq(0) }
+    end
+  end
+end

spec/workers/projects/schedule_refresh_build_artifacts_size_statistics_worker_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+RSpec.describe Projects::ScheduleRefreshBuildArtifactsSizeStatisticsWorker do
+  subject(:worker) { described_class.new }
+
+  describe '#perform' do
+    include_examples 'an idempotent worker' do
+      it 'schedules Projects::RefreshBuildArtifactsSizeStatisticsWorker to be performed with capacity' do
+        expect(Projects::RefreshBuildArtifactsSizeStatisticsWorker).to receive(:perform_with_capacity).twice
+
+        subject
+      end
+    end
+  end
+end