Makes LFS object linker process OIDs in batches

During a project import, `LfsLinkService` attempts to link `LfsObjects`
that have not already been associated with a project.  It's possible for
a large repo to have thousands of OIDs, which can cause long database
query and parsing times.  By processing a batch of 1000 at a time, we
can reduce that time at the expense of a few more SQL queries.

Closes https://gitlab.com/gitlab-org/gitlab-ce/issues/66274

app/models/lfs_object.rb

@@ -2,6 +2,7 @@
 
 class LfsObject < ApplicationRecord
   include AfterCommitQueue
+  include EachBatch
   include ObjectStorage::BackgroundMove
 
   has_many :lfs_objects_projects, dependent: :destroy # rubocop:disable Cop/ActiveRecordDependent

app/services/projects/lfs_pointers/lfs_link_service.rb

@@ -4,6 +4,8 @@
 module Projects
   module LfsPointers
     class LfsLinkService < BaseService
+      BATCH_SIZE = 1000
+
       # Accept an array of oids to link
       #
       # Returns an array with the oid of the existent lfs objects
@@ -18,16 +20,33 @@ module Projects
 
       # rubocop: disable CodeReuse/ActiveRecord
       def link_existing_lfs_objects(oids)
-        existent_lfs_objects = LfsObject.where(oid: oids)
+        all_existing_objects = []
+        iterations = 0
+
+        LfsObject.where(oid: oids).each_batch(of: BATCH_SIZE) do |existent_lfs_objects|
+          next unless existent_lfs_objects.any?
+
+          iterations += 1
+          not_linked_lfs_objects = existent_lfs_objects.where.not(id: project.all_lfs_objects)
+          project.all_lfs_objects << not_linked_lfs_objects
 
-        return [] unless existent_lfs_objects.any?
+          all_existing_objects += existent_lfs_objects.pluck(:oid)
+        end
 
-        not_linked_lfs_objects = existent_lfs_objects.where.not(id: project.all_lfs_objects)
-        project.all_lfs_objects << not_linked_lfs_objects
+        log_lfs_link_results(all_existing_objects.count, iterations)
 
-        existent_lfs_objects.pluck(:oid)
+        all_existing_objects
       end
       # rubocop: enable CodeReuse/ActiveRecord
+
+      def log_lfs_link_results(lfs_objects_linked_count, iterations)
+        Gitlab::Import::Logger.info(
+          class: self.class.name,
+          project_id: project.id,
+          project_path: project.full_path,
+          lfs_objects_linked_count: lfs_objects_linked_count,
+          iterations: iterations)
+      end
     end
   end
 end

changelogs/unreleased/sh-lfs-object-batches.yml

+---
+title: Makes LFS object linker process OIDs in batches
+merge_request: 32268
+author:
+type: performance

spec/services/projects/lfs_pointers/lfs_link_service_spec.rb

@@ -30,5 +30,23 @@ describe Projects::LfsPointers::LfsLinkService do
 
       expect(subject.execute(new_oid_list.keys)).to eq linked
     end
+
+    it 'links in batches' do
+      stub_const("#{described_class}::BATCH_SIZE", 3)
+
+      expect(Gitlab::Import::Logger)
+        .to receive(:info)
+        .with(class: described_class.name,
+              project_id: project.id,
+              project_path: project.full_path,
+              lfs_objects_linked_count: 7,
+              iterations: 3)
+
+      lfs_objects = create_list(:lfs_object, 7)
+      linked = subject.execute(lfs_objects.pluck(:oid))
+
+      expect(project.all_lfs_objects.count).to eq 9
+      expect(linked.size).to eq 7
+    end
   end
 end