Take two for MR metrics population background migration

changelogs/unreleased/41587-osw-mr-metrics-migration-cleanup.yml

@@ -0,0 +1,5 @@
+---
+title: Take two for MR metrics population background migration
+merge_request: 19097
+author:
+type: other

db/post_migrate/20180521162137_migrate_remaining_mr_metrics_populating_background_migration.rb

@@ -0,0 +1,44 @@
+class MigrateRemainingMrMetricsPopulatingBackgroundMigration < ActiveRecord::Migration
+  include Gitlab::Database::MigrationHelpers
+
+  DOWNTIME = false
+  BATCH_SIZE = 5_000
+  MIGRATION = 'PopulateMergeRequestMetricsWithEventsData'
+  DELAY_INTERVAL = 10.minutes
+
+  disable_ddl_transaction!
+
+  class MergeRequest < ActiveRecord::Base
+    self.table_name = 'merge_requests'
+
+    include ::EachBatch
+  end
+
+  def up
+    # Perform any ongoing background migration that might still be running. This
+    # avoids scheduling way too many of the same jobs on self-hosted instances
+    # if they're updating GitLab across multiple versions. The "Take one"
+    # migration was executed on 10.4 on
+    # SchedulePopulateMergeRequestMetricsWithEventsData.
+    Gitlab::BackgroundMigration.steal(MIGRATION)
+
+    metrics_not_exists_clause = <<~SQL
+      NOT EXISTS (SELECT 1 FROM merge_request_metrics
+                  WHERE merge_request_metrics.merge_request_id = merge_requests.id)
+    SQL
+
+    relation = MergeRequest.where(metrics_not_exists_clause)
+
+    # We currently have ~400_000 MR records without metrics on GitLab.com.
+    # This means it'll schedule ~80 jobs (5000 MRs each) with a 10 minutes gap,
+    # so this should take ~14 hours for all background migrations to complete.
+    #
+    queue_background_migration_jobs_by_range_at_intervals(relation,
+                                                          MIGRATION,
+                                                          DELAY_INTERVAL,
+                                                          batch_size: BATCH_SIZE)
+  end
+
+  def down
+  end
+end

lib/gitlab/import_export/import_export.yml

@@ -28,6 +28,7 @@ project_tree:
   - project_members:
     - :user
   - merge_requests:
+    - :metrics
     - notes:
       - :author
       - events:

lib/gitlab/import_export/relation_factory.rb

@@ -18,9 +18,10 @@ module Gitlab
                     label: :project_label,
                     custom_attributes: 'ProjectCustomAttribute',
                     project_badges: 'Badge',
+                    metrics: 'MergeRequest::Metrics',
                     ci_cd_settings: 'ProjectCiCdSetting' }.freeze
 
-      USER_REFERENCES = %w[author_id assignee_id updated_by_id user_id created_by_id last_edited_by_id merge_user_id resolved_by_id closed_by_id].freeze
+      USER_REFERENCES = %w[author_id assignee_id updated_by_id merged_by_id latest_closed_by_id user_id created_by_id last_edited_by_id merge_user_id resolved_by_id closed_by_id].freeze
 
       PROJECT_REFERENCES = %w[project_id source_project_id target_project_id].freeze
 
@@ -36,6 +37,15 @@ module Gitlab
         new(*args).create
       end
 
+      def self.relation_class(relation_name)
+        # There are scenarios where the model is pluralized (e.g.
+        # MergeRequest::Metrics), and we don't want to force it to singular
+        # with #classify.
+        relation_name.to_s.classify.constantize
+      rescue NameError
+        relation_name.to_s.constantize
+      end
+
       def initialize(relation_sym:, relation_hash:, members_mapper:, user:, project:, excluded_keys: [])
         @relation_name = OVERRIDES[relation_sym] || relation_sym
         @relation_hash = relation_hash.except('noteable_id')
@@ -195,7 +205,7 @@ module Gitlab
       end
 
       def relation_class
-        @relation_class ||= @relation_name.to_s.classify.constantize
+        @relation_class ||= self.class.relation_class(@relation_name)
       end
 
       def imported_object

spec/lib/gitlab/import_export/all_models.yml

@@ -310,3 +310,8 @@ lfs_file_locks:
 - user
 project_badges:
 - project
+metrics:
+- merge_request
+- latest_closed_by
+- merged_by
+- pipeline

spec/lib/gitlab/import_export/relation_factory_spec.rb

@@ -119,6 +119,25 @@ describe Gitlab::ImportExport::RelationFactory do
     end
   end
 
+  context 'overrided model with pluralized name' do
+    let(:relation_sym) { :metrics }
+
+    let(:relation_hash) do
+      {
+        'id' => 99,
+        'merge_request_id' => 99,
+        'merged_at' => Time.now,
+        'merged_by_id' => 99,
+        'latest_closed_at' => nil,
+        'latest_closed_by_id' => nil
+      }
+    end
+
+    it 'does not raise errors' do
+      expect { created_object }.not_to raise_error
+    end
+  end
+
   context 'Project references' do
     let(:relation_sym) { :project_foo_model }
     let(:relation_hash) do

spec/lib/gitlab/import_export/safe_model_attributes.yml

@@ -205,6 +205,19 @@ MergeRequestDiffFile:
 - b_mode
 - too_large
 - binary
+MergeRequest::Metrics:
+- id
+- created_at
+- updated_at
+- merge_request_id
+- pipeline_id
+- latest_closed_by_id
+- latest_closed_at
+- merged_by_id
+- merged_at
+- latest_build_started_at
+- latest_build_finished_at
+- first_deployed_to_production_at
 Ci::Pipeline:
 - id
 - project_id

spec/migrations/migrate_remaining_mr_metrics_populating_background_migration_spec.rb

@@ -0,0 +1,36 @@
+require 'spec_helper'
+require Rails.root.join('db', 'post_migrate', '20180521162137_migrate_remaining_mr_metrics_populating_background_migration.rb')
+
+describe MigrateRemainingMrMetricsPopulatingBackgroundMigration, :migration, :sidekiq do
+  let(:namespaces) { table(:namespaces) }
+  let(:projects) { table(:projects) }
+  let(:mrs) { table(:merge_requests) }
+
+  before do
+    namespaces.create!(id: 1, name: 'foo', path: 'foo')
+    projects.create!(id: 123, name: 'gitlab1', path: 'gitlab1', namespace_id: 1)
+    projects.create!(id: 456, name: 'gitlab2', path: 'gitlab2', namespace_id: 1)
+    projects.create!(id: 789, name: 'gitlab3', path: 'gitlab3', namespace_id: 1)
+    mrs.create!(title: 'foo', target_branch: 'target', source_branch: 'source', target_project_id: 123)
+    mrs.create!(title: 'bar', target_branch: 'target', source_branch: 'source', target_project_id: 456)
+    mrs.create!(title: 'kux', target_branch: 'target', source_branch: 'source', target_project_id: 789)
+  end
+
+  it 'correctly schedules background migrations' do
+    stub_const("#{described_class.name}::BATCH_SIZE", 2)
+
+    Sidekiq::Testing.fake! do
+      Timecop.freeze do
+        migrate!
+
+        expect(described_class::MIGRATION)
+          .to be_scheduled_delayed_migration(10.minutes, mrs.first.id, mrs.second.id)
+
+        expect(described_class::MIGRATION)
+          .to be_scheduled_delayed_migration(20.minutes, mrs.third.id, mrs.third.id)
+
+        expect(BackgroundMigrationWorker.jobs.size).to eq(2)
+      end
+    end
+  end
+end

spec/support/import_export/configuration_helper.rb

@@ -10,7 +10,7 @@ module ConfigurationHelper
 
   def relation_class_for_name(relation_name)
     relation_name = Gitlab::ImportExport::RelationFactory::OVERRIDES[relation_name.to_sym] || relation_name
-    relation_name.to_s.classify.constantize
+    Gitlab::ImportExport::RelationFactory.relation_class(relation_name)
   end
 
   def parsed_attributes(relation_name, attributes)