Remove N+1 queries from indexing issues

The `#as_indexed_json` method is used to convert individual documents to
the JSON that is sent to Elasticsearch during indexing. We need to
preload the data before it gets to this method. We implemented the
`.preload_indexing_data` method in
https://gitlab.com/gitlab-org/gitlab/-/merge_requests/56808 so we just
need to add this to the various classes now. This MR updates it for
`Issue` only.

We also slightly refactor the issue_instance_proxy since it's more
efficient to just use the `user_id` from the `issue_assignees` join
table than to join again to `users` and also since it wasn't possible to
get the preload to work without loading all of `users.*` which would be
wasteful for a wide table and large numbers of records.

app/models/issue.rb

@@ -447,6 +447,10 @@ class Issue < ApplicationRecord
     issue_email_participants.pluck(IssueEmailParticipant.arel_table[:email].lower)
   end
 
+  def issue_assignee_user_ids
+    issue_assignees.pluck(:user_id)
+  end
+
   private
 
   # Ensure that the metrics association is safely created and respecting the unique constraint on issue_id

ee/changelogs/unreleased/324745-preload-issues-for-elasticsearch-indexing.yml

+---
+title: Remove N+1 DB queries from indexing issues in Elasticsearch
+merge_request: 57788
+author:
+type: performance

ee/lib/elastic/latest/issue_class_proxy.rb

@@ -28,6 +28,12 @@ module Elastic
         search(query_hash, options)
       end
 
+      # rubocop: disable CodeReuse/ActiveRecord
+      def preload_indexing_data(relation)
+        relation.includes(:issue_assignees, project: [:project_feature])
+      end
+      # rubocop: enable CodeReuse/ActiveRecord
+
       private
 
       # Builds an elasticsearch query that will select documents from a

ee/lib/elastic/latest/issue_instance_proxy.rb

@@ -12,7 +12,11 @@ module Elastic
           data[attr.to_s] = safely_read_attribute_for_elasticsearch(attr)
         end
 
-        data['assignee_id'] = safely_read_attribute_for_elasticsearch(:assignee_ids)
+        # Load them through the issue_assignees table since calling
+        # assignee_ids can't be easily preloaded and does
+        # unnecessary joins
+        data['assignee_id'] = safely_read_attribute_for_elasticsearch(:issue_assignee_user_ids)
+
         data['visibility_level'] = target.project.visibility_level
         data['issues_access_level'] = safely_read_project_feature_for_elasticsearch(:issues)
 

ee/spec/services/elastic/process_bookkeeping_service_spec.rb

@@ -221,7 +221,7 @@ RSpec.describe Elastic::ProcessBookkeepingService, :clean_gitlab_redis_shared_st
 
         described_class.track!(*notes)
 
-        control = ActiveRecord::QueryRecorder.new { described_class.new.execute }
+        control = ActiveRecord::QueryRecorder.new(skip_cached: false) { described_class.new.execute }
 
         3.times do
           notes << create(:note)
@@ -234,6 +234,20 @@ RSpec.describe Elastic::ProcessBookkeepingService, :clean_gitlab_redis_shared_st
 
         expect { described_class.new.execute }.not_to exceed_all_query_limit(control)
       end
+
+      it 'does not have N+1 queries for issues' do
+        issues = create_list(:issue, 2)
+
+        described_class.track!(*issues)
+
+        control = ActiveRecord::QueryRecorder.new(skip_cached: false) { described_class.new.execute }
+
+        issues += create_list(:issue, 3)
+
+        described_class.track!(*issues)
+
+        expect { described_class.new.execute }.not_to exceed_all_query_limit(control)
+      end
     end
 
     def expect_processing(*refs, failures: [])