Track merge_requests_users usage data

To get the NSM for source code group, we need to track the number
of unique merge requests users. This will be used to determine
how much of MAUs in percentage are using merge requests.

changelogs/unreleased/216174-track-mr-usage.yml

+---
+title: Track merge_requests_users usage data
+merge_request: 32562
+author:
+type: changed

db/migrate/20200306160521_add_index_on_author_id_and_created_at_to_events.rb

@@ -4,14 +4,15 @@ class AddIndexOnAuthorIdAndCreatedAtToEvents < ActiveRecord::Migration[6.0]
   include Gitlab::Database::MigrationHelpers
 
   DOWNTIME = false
+  INDEX_NAME = 'index_events_on_author_id_and_created_at'
 
   disable_ddl_transaction!
 
   def up
-    add_concurrent_index :events, [:author_id, :created_at]
+    add_concurrent_index :events, [:author_id, :created_at], name: INDEX_NAME
   end
 
   def down
-    remove_concurrent_index :events, [:author_id, :created_at]
+    remove_concurrent_index :events, INDEX_NAME
   end
 end

db/post_migrate/20200528123703_add_merge_request_partial_index_to_events.rb

+# frozen_string_literal: true
+
+class AddMergeRequestPartialIndexToEvents < ActiveRecord::Migration[6.0]
+  include Gitlab::Database::MigrationHelpers
+
+  DOWNTIME = false
+
+  disable_ddl_transaction!
+
+  INDEX_NAME = 'index_events_on_author_id_and_created_at_merge_requests'
+
+  def up
+    add_concurrent_index(
+      :events,
+      [:author_id, :created_at],
+      name: INDEX_NAME,
+      where: "(target_type = 'MergeRequest')"
+    )
+  end
+
+  def down
+    remove_concurrent_index :events, INDEX_NAME
+  end
+end

db/structure.sql

@@ -9706,6 +9706,8 @@ CREATE INDEX index_events_on_action ON public.events USING btree (action);
 
 CREATE INDEX index_events_on_author_id_and_created_at ON public.events USING btree (author_id, created_at);
 
+CREATE INDEX index_events_on_author_id_and_created_at_merge_requests ON public.events USING btree (author_id, created_at) WHERE ((target_type)::text = 'MergeRequest'::text);
+
 CREATE INDEX index_events_on_author_id_and_project_id ON public.events USING btree (author_id, project_id);
 
 CREATE INDEX index_events_on_group_id_partial ON public.events USING btree (group_id) WHERE (group_id IS NOT NULL);
@@ -14036,5 +14038,6 @@ COPY "schema_migrations" (version) FROM STDIN;
 20200527151413
 20200527152116
 20200527152657
+20200528123703
 \.
 

ee/lib/ee/gitlab/usage_data.rb

@@ -39,8 +39,8 @@ module EE
 
         override :uncached_data
         def uncached_data
-          time_period = { created_at: 28.days.ago..Time.current }
-          usage_activity_by_stage_monthly = usage_activity_by_stage(:usage_activity_by_stage_monthly, time_period)
+          usage_activity_by_stage_monthly = usage_activity_by_stage(:usage_activity_by_stage_monthly, default_time_period)
+
           super
             .merge(usage_activity_by_stage)
             .merge(usage_activity_by_stage_monthly)

lib/gitlab/usage_data.rb

@@ -32,6 +32,7 @@ module Gitlab
           .merge(cycle_analytics_usage_data)
           .merge(object_store_usage_data)
           .merge(recording_ce_finish_data)
+          .merge(merge_requests_usage_data(default_time_period))
       end
 
       def to_json(force_refresh: false)
@@ -382,6 +383,27 @@ module Gitlab
         {} # augmented in EE
       end
 
+      # rubocop: disable CodeReuse/ActiveRecord
+      def merge_requests_usage_data(time_period)
+        query =
+          Event
+            .where(target_type: Event::TARGET_TYPES[:merge_request].to_s)
+            .where(time_period)
+
+        merge_request_users = distinct_count(
+          query,
+          :author_id,
+          batch_size: 5_000, # Based on query performance, this is the optimal batch size.
+          start: User.minimum(:id),
+          finish: User.maximum(:id)
+        )
+
+        {
+          merge_requests_users: merge_request_users
+        }
+      end
+      # rubocop: enable CodeReuse/ActiveRecord
+
       def installation_type
         if Rails.env.production?
           Gitlab::INSTALLATION_TYPE
@@ -389,6 +411,10 @@ module Gitlab
           "gitlab-development-kit"
         end
       end
+
+      def default_time_period
+        { created_at: 28.days.ago..Time.current }
+      end
     end
   end
 end

lib/gitlab/utils/usage_data.rb

@@ -49,9 +49,9 @@ module Gitlab
         FALLBACK
       end
 
-      def distinct_count(relation, column = nil, batch: true, start: nil, finish: nil)
+      def distinct_count(relation, column = nil, batch: true, batch_size: nil, start: nil, finish: nil)
         if batch && Feature.enabled?(:usage_ping_batch_counter, default_enabled: true)
-          Gitlab::Database::BatchCount.batch_distinct_count(relation, column, start: start, finish: finish)
+          Gitlab::Database::BatchCount.batch_distinct_count(relation, column, batch_size: batch_size, start: start, finish: finish)
         else
           relation.distinct_count_by(column)
         end

spec/lib/gitlab/usage_data_spec.rb

@@ -574,4 +574,27 @@ describe Gitlab::UsageData, :aggregate_failures do
 
     it_behaves_like 'usage data execution'
   end
+
+  describe '#merge_requests_usage_data' do
+    let(:time_period) { { created_at: 2.days.ago..Time.current } }
+    let(:merge_request) { create(:merge_request) }
+    let(:other_user) { create(:user) }
+    let(:another_user) { create(:user) }
+
+    before do
+      create(:event, target: merge_request, author: merge_request.author, created_at: 1.day.ago)
+      create(:event, target: merge_request, author: merge_request.author, created_at: 1.hour.ago)
+      create(:event, target: merge_request, author: merge_request.author, created_at: 3.days.ago)
+      create(:event, target: merge_request, author: other_user, created_at: 1.day.ago)
+      create(:event, target: merge_request, author: other_user, created_at: 1.hour.ago)
+      create(:event, target: merge_request, author: other_user, created_at: 3.days.ago)
+      create(:event, target: merge_request, author: another_user, created_at: 4.days.ago)
+    end
+
+    it 'returns the distinct count of users using merge requests (via events table) within the specified time period' do
+      expect(described_class.merge_requests_usage_data(time_period)).to eq(
+        merge_requests_users: 2
+      )
+    end
+  end
 end