LFS export records repository_type data

A project can have the same `LfsObject` linked with up to three
`LfsObjectsProject` records. Each of these records would be for a
different repository, recorded in the `repository_type` property. The
different repositories at time of writing are "project", "wiki", and
"design". See https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/13894

This change exports the list of `repository_type`s as a JSON mapping of
oid => repository_types, which are imported to recreate the correct
`LfsObjectsProject` records.

https://gitlab.com/gitlab-org/gitlab-ee/issues/11090

changelogs/unreleased/11090-export-design-management-lfs-data.yml

+---
+title: Add support for exporting repository type data for LFS objects
+merge_request: 30830
+author:
+type: changed

doc/administration/logs.md

@@ -283,9 +283,6 @@ Introduced in GitLab 11.3. This file lives in `/var/log/gitlab/gitlab-rails/impo
 Omnibus GitLab packages or in `/home/git/gitlab/log/importer.log` for
 installations from source.
 
-Currently it logs the progress of project imports from the Bitbucket Server
-importer. Future importers may use this file.
-
 ## `auth.log`
 
 Introduced in GitLab 12.0. This file lives in `/var/log/gitlab/gitlab-rails/auth.log` for

doc/development/import_export.md

@@ -19,6 +19,7 @@ Project.find_by_full_path('group/project').import_state.slice(:jid, :status, :la
 grep JID /var/log/gitlab/sidekiq/current
 grep "Import/Export error" /var/log/gitlab/sidekiq/current
 grep "Import/Export backtrace" /var/log/gitlab/sidekiq/current
+tail /var/log/gitlab/gitlab-rails/importer.log
 ```
 
 ## Troubleshooting performance issues

lib/gitlab/import_export.rb

@@ -4,7 +4,9 @@ module Gitlab
   module ImportExport
     extend self
 
-    # For every version update, the version history in import_export.md has to be kept up to date.
+    # For every version update the version history in these docs must be kept up to date:
+    # - development/import_export.md
+    # - user/project/settings/import_export.md
     VERSION = '0.2.4'.freeze
     FILENAME_LIMIT = 50
 
@@ -28,6 +30,14 @@ module Gitlab
       "project.bundle"
     end
 
+    def lfs_objects_filename
+      "lfs-objects.json"
+    end
+
+    def lfs_objects_storage
+      "lfs-objects"
+    end
+
     def config_file
       Rails.root.join('lib/gitlab/import_export/import_export.yml')
     end

lib/gitlab/import_export/lfs_restorer.rb

@@ -3,6 +3,10 @@
 module Gitlab
   module ImportExport
     class LfsRestorer
+      include Gitlab::Utils::StrongMemoize
+
+      attr_accessor :project, :shared
+
       def initialize(project:, shared:)
         @project = project
         @shared = shared
@@ -17,7 +21,7 @@ module Gitlab
 
         true
       rescue => e
-        @shared.error(e)
+        shared.error(e)
         false
       end
 
@@ -29,16 +33,57 @@ module Gitlab
 
         lfs_object = LfsObject.find_or_initialize_by(oid: oid, size: size)
         lfs_object.file = File.open(path) unless lfs_object.file&.exists?
+        lfs_object.save! if lfs_object.changed?
+
+        repository_types(oid).each do |repository_type|
+          LfsObjectsProject.create!(
+            project: project,
+            lfs_object: lfs_object,
+            repository_type: repository_type
+          )
+        end
+      end
+
+      def repository_types(oid)
+        # We allow support for imports created before the `lfs-objects.json`
+        # file was generated. In this case, the restorer will link an LFS object
+        # with a single `lfs_objects_projects` relation.
+        #
+        # This allows us backwards-compatibility without version bumping.
+        # See https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/30830#note_192608870
+        return ['project'] unless has_lfs_json?
 
-        @project.all_lfs_objects << lfs_object
+        lfs_json[oid]
       end
 
       def lfs_file_paths
         @lfs_file_paths ||= Dir.glob("#{lfs_storage_path}/*")
       end
 
+      def has_lfs_json?
+        strong_memoize(:has_lfs_json) do
+          File.exist?(lfs_json_path)
+        end
+      end
+
+      def lfs_json
+        return {} unless has_lfs_json?
+
+        @lfs_json ||=
+          begin
+            json = IO.read(lfs_json_path)
+            ActiveSupport::JSON.decode(json)
+          rescue
+            raise Gitlab::ImportExport::Error.new('Incorrect JSON format')
+          end
+      end
+
       def lfs_storage_path
-        File.join(@shared.export_path, 'lfs-objects')
+        File.join(shared.export_path, ImportExport.lfs_objects_storage)
+      end
+
+      def lfs_json_path
+        File.join(shared.export_path, ImportExport.lfs_objects_filename)
       end
     end
   end

lib/gitlab/import_export/lfs_saver.rb

@@ -5,25 +5,40 @@ module Gitlab
     class LfsSaver
       include Gitlab::ImportExport::CommandLineUtil
 
+      attr_accessor :lfs_json, :project, :shared
+
+      BATCH_SIZE = 100
+
       def initialize(project:, shared:)
         @project = project
         @shared = shared
+        @lfs_json = {}
       end
 
       def save
-        @project.all_lfs_objects.each do |lfs_object|
+        project.all_lfs_objects.find_in_batches(batch_size: BATCH_SIZE) do |batch|
+          batch.each do |lfs_object|
             save_lfs_object(lfs_object)
           end
 
+          append_lfs_json_for_batch(batch) if write_lfs_json_enabled?
+        end
+
+        write_lfs_json if write_lfs_json_enabled?
+
         true
       rescue => e
-        @shared.error(e)
+        shared.error(e)
 
         false
       end
 
       private
 
+      def write_lfs_json_enabled?
+        ::Feature.enabled?(:export_lfs_objects_projects, default_enabled: true)
+      end
+
       def save_lfs_object(lfs_object)
         if lfs_object.local_store?
           copy_file_for_lfs_object(lfs_object)
@@ -45,12 +60,36 @@ module Gitlab
         copy_files(lfs_object.file.path, destination_path_for_object(lfs_object))
       end
 
+      def append_lfs_json_for_batch(lfs_objects_batch)
+        lfs_objects_projects = LfsObjectsProject
+                                .select('lfs_objects.oid, array_agg(distinct lfs_objects_projects.repository_type) as repository_types')
+                                .joins(:lfs_object)
+                                .where(project: project, lfs_object: lfs_objects_batch)
+                                .group('lfs_objects.oid')
+
+        lfs_objects_projects.each do |group|
+          oid = group.oid
+
+          lfs_json[oid] ||= []
+          lfs_json[oid] += group.repository_types
+        end
+      end
+
+      def write_lfs_json
+        mkdir_p(shared.export_path)
+        File.write(lfs_json_path, lfs_json.to_json)
+      end
+
       def destination_path_for_object(lfs_object)
         File.join(lfs_export_path, lfs_object.oid)
       end
 
       def lfs_export_path
-        File.join(@shared.export_path, 'lfs-objects')
+        File.join(shared.export_path, ImportExport.lfs_objects_storage)
+      end
+
+      def lfs_json_path
+        File.join(shared.export_path, ImportExport.lfs_objects_filename)
       end
     end
   end

spec/factories/lfs_objects.rb

@@ -14,6 +14,7 @@ FactoryBot.define do
   # objects, so the test needs to decide which (if any) object gets it
   trait :correct_oid do
     oid 'b804383982bb89b00e828e3f44c038cc991d3d1768009fc39ba8e2c081b9fb75'
+    size 1062
   end
 
   trait :object_storage do

spec/lib/gitlab/import_export/lfs_restorer_spec.rb

@@ -6,6 +6,7 @@ describe Gitlab::ImportExport::LfsRestorer do
   let(:export_path) { "#{Dir.tmpdir}/lfs_object_restorer_spec" }
   let(:project) { create(:project) }
   let(:shared) { project.import_export_shared }
+  let(:saver) { Gitlab::ImportExport::LfsSaver.new(project: project, shared: shared) }
   subject(:restorer) { described_class.new(project: project, shared: shared) }
 
   before do
@@ -19,43 +20,69 @@ describe Gitlab::ImportExport::LfsRestorer do
 
   describe '#restore' do
     context 'when the archive contains lfs files' do
-      let(:dummy_lfs_file_path) { File.join(shared.export_path, 'lfs-objects', 'dummy') }
+      let(:lfs_object) { create(:lfs_object, :correct_oid, :with_file) }
+
+      # Use the LfsSaver to save data to be restored
+      def save_lfs_data
+        %w(project wiki).each do |repository_type|
+          create(
+            :lfs_objects_project,
+            project: project,
+            repository_type: repository_type,
+            lfs_object: lfs_object
+          )
+        end
+
+        saver.save
 
-      def create_lfs_object_with_content(content)
-        dummy_lfs_file = Tempfile.new('existing')
-        File.write(dummy_lfs_file.path, content)
-        size = dummy_lfs_file.size
-        oid = LfsObject.calculate_oid(dummy_lfs_file.path)
-        LfsObject.create!(oid: oid, size: size, file: dummy_lfs_file)
+        project.lfs_objects.delete_all
       end
 
       before do
-        FileUtils.mkdir_p(File.dirname(dummy_lfs_file_path))
-        File.write(dummy_lfs_file_path, 'not very large')
-        allow(restorer).to receive(:lfs_file_paths).and_return([dummy_lfs_file_path])
+        save_lfs_data
+        project.reload
       end
 
-      it 'creates an lfs object for the project' do
-        expect { restorer.restore }.to change { project.reload.lfs_objects.size }.by(1)
+      it 'succeeds' do
+        expect(restorer.restore).to eq(true)
+        expect(shared.errors).to be_empty
       end
 
-      it 'assigns the file correctly' do
+      it 'does not create a new `LfsObject` records, as one already exists' do
+        expect { restorer.restore }.not_to change { LfsObject.count }
+      end
+
+      it 'creates new `LfsObjectsProject` records in order to link the project to the existing `LfsObject`' do
+        expect { restorer.restore }.to change { LfsObjectsProject.count }.by(2)
+      end
+
+      it 'restores the correct `LfsObject` records' do
         restorer.restore
 
-        expect(project.lfs_objects.first.file.read).to eq('not very large')
+        expect(project.lfs_objects).to contain_exactly(lfs_object)
       end
 
-      it 'links an existing LFS object if it existed' do
-        lfs_object = create_lfs_object_with_content('not very large')
+      it 'restores the correct `LfsObjectsProject` records for the project' do
+        restorer.restore
+
+        expect(
+          project.lfs_objects_projects.pluck(:repository_type)
+        ).to contain_exactly('project', 'wiki')
+      end
 
+      it 'assigns the file correctly' do
         restorer.restore
 
-        expect(project.lfs_objects).to include(lfs_object)
+        expect(project.lfs_objects.first.file.read).to eq(lfs_object.file.read)
       end
 
-      it 'succeeds' do
-        expect(restorer.restore).to be_truthy
-        expect(shared.errors).to be_empty
+      context 'when there is not an existing `LfsObject`' do
+        before do
+          lfs_object.destroy
+        end
+
+        it 'creates a new lfs object' do
+          expect { restorer.restore }.to change { LfsObject.count }.by(1)
         end
 
         it 'stores the upload' do
@@ -65,6 +92,29 @@ describe Gitlab::ImportExport::LfsRestorer do
         end
       end
 
+      context 'when there is no lfs-objects.json file' do
+        before do
+          json_file = File.join(shared.export_path, ::Gitlab::ImportExport.lfs_objects_filename)
+
+          FileUtils.rm_rf(json_file)
+        end
+
+        it 'restores the correct `LfsObject` records' do
+          restorer.restore
+
+          expect(project.lfs_objects).to contain_exactly(lfs_object)
+        end
+
+        it 'restores a single `LfsObjectsProject` record for the project with "project" for the `repository_type`' do
+          restorer.restore
+
+          expect(
+            project.lfs_objects_projects.pluck(:repository_type)
+          ).to contain_exactly('project')
+        end
+      end
+    end
+
     context 'without any LFS-objects' do
       it 'succeeds' do
         expect(restorer.restore).to be_truthy

spec/lib/gitlab/import_export/lfs_saver_spec.rb

@@ -19,6 +19,11 @@ describe Gitlab::ImportExport::LfsSaver do
   describe '#save' do
     context 'when the project has LFS objects locally stored' do
       let(:lfs_object) { create(:lfs_object, :with_file) }
+      let(:lfs_json_file) { File.join(shared.export_path, Gitlab::ImportExport.lfs_objects_filename) }
+
+      def lfs_json
+        JSON.parse(IO.read(lfs_json_file))
+      end
 
       before do
         project.lfs_objects << lfs_object
@@ -35,6 +40,45 @@ describe Gitlab::ImportExport::LfsSaver do
 
         expect(File).to exist("#{shared.export_path}/lfs-objects/#{lfs_object.oid}")
       end
+
+      describe 'saving a json file' do
+        before do
+          # Create two more LfsObjectProject records with different `repository_type`s
+          %w(wiki design).each do |repository_type|
+            create(
+              :lfs_objects_project,
+              project: project,
+              repository_type: repository_type,
+              lfs_object: lfs_object
+            )
+          end
+
+          FileUtils.rm_rf(lfs_json_file)
+        end
+
+        it 'saves a json file correctly' do
+          saver.save
+
+          expect(File.exist?(lfs_json_file)).to eq(true)
+          expect(lfs_json).to eq(
+            {
+              lfs_object.oid => [
+                LfsObjectsProject.repository_types['wiki'],
+                LfsObjectsProject.repository_types['design'],
+                nil
+              ]
+            }
+          )
+        end
+
+        it 'does not save a json file if feature is disabled' do
+          stub_feature_flags(export_lfs_objects_projects: false)
+
+          saver.save
+
+          expect(File.exist?(lfs_json_file)).to eq(false)
+        end
+      end
     end
 
     context 'when the LFS objects are stored in object storage' do
@@ -42,8 +86,11 @@ describe Gitlab::ImportExport::LfsSaver do
 
       before do
         allow(LfsObjectUploader).to receive(:object_store_enabled?).and_return(true)
-        allow(lfs_object.file).to receive(:url).and_return('http://my-object-storage.local')
         project.lfs_objects << lfs_object
+
+        expect_next_instance_of(LfsObjectUploader) do |instance|
+          expect(instance).to receive(:url).and_return('http://my-object-storage.local')
+        end
       end
 
       it 'downloads the file to include in an archive' do