Add latest changes from gitlab-org/gitlab@16-0-stable-eev16.0.0-rc42

1 files changed, 66 insertions, 0 deletions

diff --git a/lib/gitlab/background_migration/issues_internal_id_scope_updater.rb b/lib/gitlab/background_migration/issues_internal_id_scope_updater.rb
new file mode 100644
index 00000000000..21ca4392003
--- /dev/null
+++ b/lib/gitlab/background_migration/issues_internal_id_scope_updater.rb
@@ -0,0 +1,66 @@
+# frozen_string_literal: true
+
+module Gitlab
+  module BackgroundMigration
+    # Migrates internal_ids records for `usage: issues` from project to namespace scope.
+    # For project issues it will be project namespace, for group issues it will be group namespace.
+    class IssuesInternalIdScopeUpdater < ::Gitlab::BackgroundMigration::BatchedMigrationJob
+      operation_name :issues_internal_id_scope_updater
+      feature_category :database
+
+      ISSUES_USAGE = 0 # see Enums::InternalId#usage_resources[:issues]
+
+      scope_to ->(relation) do
+        relation.where(usage: ISSUES_USAGE).where.not(project_id: nil)
+      end
+
+      def perform
+        each_sub_batch do |sub_batch|
+          create_namespace_scoped_records(sub_batch)
+          delete_project_scoped_records(sub_batch)
+        end
+      end
+
+      private
+
+      def delete_project_scoped_records(sub_batch)
+        # There is no need to keep the project scoped issues usage as we move to scoping issues to namespace.
+        # Also in case we do decide to move back to scoping issues usage to project, we are better off if the
+        # project record is not present as that would result in overlapping IIDs because project scoped issues
+        # usage will have outdated IIDs left in the DB
+        log_info("Deleted internal_ids records", ids: sub_batch.pluck(:id))
+
+        connection.execute(
+          <<~SQL
+            DELETE FROM internal_ids WHERE id IN (#{sub_batch.select(:id).to_sql})
+          SQL
+        )
+      end
+
+      def create_namespace_scoped_records(sub_batch)
+        # Creates a corresponding namespace scoped record for every `issues` usage scoped to a project.
+        # On conflict it means the record was already created when a new issue is created with the
+        # newly namespace scoped Issue model, see Issue#has_internal_id definition. In which case to
+        # make sure we have the namespace_id scoped record set to the greatest of the two last_values.
+        created_records_ids = connection.execute(
+          <<~SQL
+              INSERT INTO internal_ids (usage, last_value, namespace_id)
+                SELECT #{ISSUES_USAGE}, last_value, project_namespace_id
+                FROM internal_ids
+                INNER JOIN projects ON projects.id = internal_ids.project_id
+                WHERE internal_ids.id IN(#{sub_batch.select(:id).to_sql})
+              ON CONFLICT (usage, namespace_id) WHERE namespace_id IS NOT NULL
+              DO UPDATE SET last_value = GREATEST(EXCLUDED.last_value, internal_ids.last_value)
+              RETURNING id;
+          SQL
+        )
+
+        log_info("Created/updated internal_ids records", ids: created_records_ids.field_values('id'))
+      end
+
+      def log_info(message, **extra)
+        ::Gitlab::BackgroundMigration::Logger.info(migrator: self.class.to_s, message: message, **extra)
+      end
+    end
+  end
+end