Support LFS objects when creating a project by import

app/models/repository.rb

@@ -956,6 +956,10 @@ class Repository
     blob_data_at(sha, path)
   end
+  def lfsconfig_for(sha)
+    blob_data_at(sha, '.lfsconfig')
+  end
+
   def fetch_ref(source_repository, source_ref:, target_ref:)
     raw_repository.fetch_ref(source_repository.raw_repository, source_ref: source_ref, target_ref: target_ref)
   end

app/services/base_service.rb

@@ -3,7 +3,7 @@ class BaseService
   attr_accessor :project, :current_user, :params
-  def initialize(project, user, params = {})
+  def initialize(project, user = nil, params = {})
     @project, @current_user, @params = project, user, params.dup
   end

app/services/projects/import_service.rb

@@ -17,6 +17,8 @@ module Projects
     def execute
       add_repository_to_project
+      download_lfs_objects
+
       import_data
       success
@@ -37,7 +39,7 @@ module Projects
       # We should skip the repository for a GitHub import or GitLab project import,
       # because these importers fetch the project repositories for us.
-      return if has_importer? && importer_class.try(:imports_repository?)
+      return if importer_imports_repository?
       if unknown_url?
         # In this case, we only want to import issues, not a repository.
@@ -73,6 +75,27 @@ module Projects
       end
     end
+    def download_lfs_objects
+      # In this case, we only want to import issues
+      return if unknown_url?
+
+      # If it has its own repository importer, it has to implements its own lfs import download
+      return if importer_imports_repository?
+
+      return unless project.lfs_enabled?
+
+      oids_to_download = Projects::LfsPointers::LfsImportService.new(project).execute
+      download_service = Projects::LfsPointers::LfsDownloadService.new(project)
+
+      oids_to_download.each do |oid, link|
+        download_service.execute(oid, link)
+      end
+    rescue => e
+      # Right now, to avoid aborting the importing process, we silently fail
+      # if any exception raises.
+      Rails.logger.error("The Lfs import process failed. #{e.message}")
+    end
+
     def import_data
       return unless has_importer?
@@ -98,5 +121,9 @@ module Projects
     def unknown_url?
       project.import_url == Project::UNKNOWN_IMPORT_URL
     end
+
+    def importer_imports_repository?
+      has_importer? && importer_class.try(:imports_repository?)
+    end
   end
 end

app/services/projects/lfs_pointers/lfs_download_link_list_service.rb

+# This service lists the download link from a remote source based on the
+# oids provided
+module Projects
+  module LfsPointers
+    class LfsDownloadLinkListService < BaseService
+      DOWNLOAD_ACTION = 'download'.freeze
+
+      DownloadLinksError = Class.new(StandardError)
+      DownloadLinkNotFound = Class.new(StandardError)
+
+      attr_reader :remote_uri
+
+      def initialize(project, remote_uri: nil)
+        super(project)
+
+        @remote_uri = remote_uri
+      end
+
+      # This method accepts two parameters:
+      # - oids: hash of oids to query. The structure is { lfs_file_oid => lfs_file_size }
+      #
+      # Returns a hash with the structure { lfs_file_oids => download_link }
+      def execute(oids)
+        return {} unless project&.lfs_enabled? && remote_uri && oids.present?
+
+        get_download_links(oids)
+      end
+
+      private
+
+      def get_download_links(oids)
+        response = Gitlab::HTTP.post(remote_uri,
+                                     body: request_body(oids),
+                                     headers: headers)
+
+        raise DownloadLinksError, response.message unless response.success?
+
+        parse_response_links(response['objects'])
+      end
+
+      def parse_response_links(objects_response)
+        objects_response.each_with_object({}) do |entry, link_list|
+          begin
+            oid = entry['oid']
+            link = entry.dig('actions', DOWNLOAD_ACTION, 'href')
+
+            raise DownloadLinkNotFound unless link
+
+            link_list[oid] = add_credentials(link)
+          rescue DownloadLinkNotFound, URI::InvalidURIError
+            Rails.logger.error("Link for Lfs Object with oid #{oid} not found or invalid.")
+          end
+        end
+      end
+
+      def request_body(oids)
+        {
+          operation: DOWNLOAD_ACTION,
+          objects: oids.map { |oid, size| { oid: oid, size: size } }
+        }.to_json
+      end
+
+      def headers
+        {
+          'Accept' => LfsRequest::CONTENT_TYPE,
+          'Content-Type' => LfsRequest::CONTENT_TYPE
+        }.freeze
+      end
+
+      def add_credentials(link)
+        uri = URI.parse(link)
+
+        if should_add_credentials?(uri)
+          uri.user = remote_uri.user
+          uri.password = remote_uri.password
+        end
+
+        uri.to_s
+      end
+
+      # The download link can be a local url or an object storage url
+      # If the download link has the some host as the import url then
+      # we add the same credentials because we may need them
+      def should_add_credentials?(link_uri)
+        url_credentials? && link_uri.host == remote_uri.host
+      end
+
+      def url_credentials?
+        remote_uri.user.present? || remote_uri.password.present?
+      end
+    end
+  end
+end

app/services/projects/lfs_pointers/lfs_download_service.rb

+# This service downloads and links lfs objects from a remote URL
+module Projects
+  module LfsPointers
+    class LfsDownloadService < BaseService
+      def execute(oid, url)
+        return unless project&.lfs_enabled? && oid.present? && url.present?
+
+        return if LfsObject.exists?(oid: oid)
+
+        sanitized_uri = Gitlab::UrlSanitizer.new(url)
+
+        with_tmp_file(oid) do |file|
+          size = download_and_save_file(file, sanitized_uri)
+          lfs_object = LfsObject.new(oid: oid, size: size, file: file)
+
+          project.all_lfs_objects << lfs_object
+        end
+      rescue StandardError => e
+        Rails.logger.error("LFS file with oid #{oid} could't be downloaded from #{sanitized_uri.sanitized_url}: #{e.message}")
+      end
+
+      private
+
+      def download_and_save_file(file, sanitized_uri)
+        IO.copy_stream(open(sanitized_uri.sanitized_url, headers(sanitized_uri)), file)
+      end
+
+      def headers(sanitized_uri)
+        {}.tap do |headers|
+          credentials = sanitized_uri.credentials
+
+          if credentials[:user].present? || credentials[:password].present?
+            # Using authentication headers in the request
+            headers[:http_basic_authentication] = [credentials[:user], credentials[:password]]
+          end
+        end
+      end
+
+      def with_tmp_file(oid)
+        create_tmp_storage_dir
+
+        File.open(File.join(tmp_storage_dir, oid), 'w') { |file| yield file }
+      end
+
+      def create_tmp_storage_dir
+        FileUtils.makedirs(tmp_storage_dir) unless Dir.exist?(tmp_storage_dir)
+      end
+
+      def tmp_storage_dir
+        @tmp_storage_dir ||= File.join(storage_dir, 'tmp', 'download')
+      end
+
+      def storage_dir
+        @storage_dir ||= Gitlab.config.lfs.storage_path
+      end
+    end
+  end
+end

app/services/projects/lfs_pointers/lfs_import_service.rb

+# This service manages the whole worflow of discovering the Lfs files in a
+# repository, linking them to the project and downloading (and linking) the non
+# existent ones.
+module Projects
+  module LfsPointers
+    class LfsImportService < BaseService
+      include Gitlab::Utils::StrongMemoize
+
+      HEAD_REV = 'HEAD'.freeze
+      LFS_ENDPOINT_PATTERN = /^\t?url\s*=\s*(.+)$/.freeze
+      LFS_BATCH_API_ENDPOINT = '/info/lfs/objects/batch'.freeze
+
+      LfsImportError = Class.new(StandardError)
+
+      def execute
+        return {} unless project&.lfs_enabled?
+
+        if external_lfs_endpoint?
+          # If the endpoint host is different from the import_url it means
+          # that the repo is using a third party service for storing the LFS files.
+          # In this case, we have to disable lfs in the project
+          disable_lfs!
+
+          return {}
+        end
+
+        get_download_links
+      rescue LfsDownloadLinkListService::DownloadLinksError => e
+        raise LfsImportError, "The LFS objects download list couldn't be imported. Error: #{e.message}"
+      end
+
+      private
+
+      def external_lfs_endpoint?
+        lfsconfig_endpoint_uri && lfsconfig_endpoint_uri.host != import_uri.host
+      end
+
+      def disable_lfs!
+        project.update(lfs_enabled: false)
+      end
+
+      def get_download_links
+        existent_lfs = LfsListService.new(project).execute
+        linked_oids = LfsLinkService.new(project).execute(existent_lfs.keys)
+
+        # Retrieving those oids not linked and which we need to download
+        not_linked_lfs = existent_lfs.except(*linked_oids)
+
+        LfsDownloadLinkListService.new(project, remote_uri: current_endpoint_uri).execute(not_linked_lfs)
+      end
+
+      def lfsconfig_endpoint_uri
+        strong_memoize(:lfsconfig_endpoint_uri) do
+          # Retrieveing the blob data from the .lfsconfig file
+          data = project.repository.lfsconfig_for(HEAD_REV)
+          # Parsing the data to retrieve the url
+          parsed_data = data&.match(LFS_ENDPOINT_PATTERN)
+
+          if parsed_data
+            URI.parse(parsed_data[1]).tap do |endpoint|
+              endpoint.user ||= import_uri.user
+              endpoint.password ||= import_uri.password
+            end
+          end
+        end
+      rescue URI::InvalidURIError
+        raise LfsImportError, 'Invalid URL in .lfsconfig file'
+      end
+
+      def import_uri
+        @import_uri ||= URI.parse(project.import_url)
+      rescue URI::InvalidURIError
+        raise LfsImportError, 'Invalid project import URL'
+      end
+
+      def current_endpoint_uri
+        (lfsconfig_endpoint_uri || default_endpoint_uri)
+      end
+
+      # The import url must end with '.git' here we ensure it is
+      def default_endpoint_uri
+        @default_endpoint_uri ||= begin
+          import_uri.dup.tap do |uri|
+            path = uri.path.gsub(%r(/$), '')
+            path += '.git' unless path.ends_with?('.git')
+            uri.path = path + LFS_BATCH_API_ENDPOINT
+          end
+        end
+      end
+    end
+  end
+end

app/services/projects/lfs_pointers/lfs_link_service.rb

+# Given a list of oids, this services links the existent Lfs Objects to the project
+module Projects
+  module LfsPointers
+    class LfsLinkService < BaseService
+      # Accept an array of oids to link
+      #
+      # Returns a hash with the same structure with oids linked
+      def execute(oids)
+        return {} unless project&.lfs_enabled?
+
+        # Search and link existing LFS Object
+        link_existing_lfs_objects(oids)
+      end
+
+      private
+
+      def link_existing_lfs_objects(oids)
+        existent_lfs_objects = LfsObject.where(oid: oids)
+
+        return [] unless existent_lfs_objects.any?
+
+        not_linked_lfs_objects = existent_lfs_objects.where.not(id: project.all_lfs_objects)
+        project.all_lfs_objects << not_linked_lfs_objects
+
+        existent_lfs_objects.pluck(:oid)
+      end
+    end
+  end
+end

app/services/projects/lfs_pointers/lfs_list_service.rb

+# This service list all existent Lfs objects in a repository
+module Projects
+  module LfsPointers
+    class LfsListService < BaseService
+      REV = 'HEAD'.freeze
+
+      # Retrieve all lfs blob pointers and returns a hash
+      # with the structure { lfs_file_oid => lfs_file_size }
+      def execute
+        return {} unless project&.lfs_enabled?
+
+        Gitlab::Git::LfsChanges.new(project.repository, REV)
+                               .all_pointers
+                               .map! { |blob| [blob.lfs_oid, blob.lfs_size] }
+                               .to_h
+      end
+    end
+  end
+end

app/workers/all_queues.yml

@@ -31,12 +31,14 @@
 - github_importer:github_import_import_diff_note
 - github_importer:github_import_import_issue
 - github_importer:github_import_import_note
+- github_importer:github_import_import_lfs_object
 - github_importer:github_import_import_pull_request
 - github_importer:github_import_refresh_import_jid
 - github_importer:github_import_stage_finish_import
 - github_importer:github_import_stage_import_base_data
 - github_importer:github_import_stage_import_issues_and_diff_notes
 - github_importer:github_import_stage_import_notes
+- github_importer:github_import_stage_import_lfs_objects
 - github_importer:github_import_stage_import_pull_requests
 - github_importer:github_import_stage_import_repository

app/workers/gitlab/github_import/advance_stage_worker.rb

@@ -21,6 +21,7 @@ module Gitlab
       STAGES = {
         issues_and_diff_notes: Stage::ImportIssuesAndDiffNotesWorker,
         notes: Stage::ImportNotesWorker,
+        lfs_objects: Stage::ImportLfsObjectsWorker,
         finish: Stage::FinishImportWorker
       }.freeze

app/workers/gitlab/github_import/import_lfs_object_worker.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module GithubImport
+    class ImportLfsObjectWorker
+      include ObjectImporter
+
+      def representation_class
+        Representation::LfsObject
+      end
+
+      def importer_class
+        Importer::LfsObjectImporter
+      end
+
+      def counter_name
+        :github_importer_imported_lfs_objects
+      end
+
+      def counter_description
+        'The number of imported GitHub Lfs Objects'
+      end
+    end
+  end
+end

app/workers/gitlab/github_import/stage/import_lfs_objects_worker.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module GithubImport
+    module Stage
+      class ImportLfsObjectsWorker
+        include ApplicationWorker
+        include GithubImport::Queue
+        include StageMethods
+
+        def perform(project_id)
+          return unless (project = find_project(project_id))
+
+          import(project)
+        end
+
+        # project - An instance of Project.
+        def import(project)
+          waiter = Importer::LfsObjectsImporter
+            .new(project, nil)
+            .execute
+
+          AdvanceStageWorker.perform_async(
+            project.id,
+            { waiter.key => waiter.jobs_remaining },
+            :finish
+          )
+        end
+      end
+    end
+  end
+end

app/workers/gitlab/github_import/stage/import_notes_worker.rb

@@ -18,7 +18,7 @@ module Gitlab
           AdvanceStageWorker.perform_async(
             project.id,
             { waiter.key => waiter.jobs_remaining },
-            :finish
+            :lfs_objects
           )
         end
       end

changelogs/unreleased/fj-40401-support-import-lfs-objects.yml

+---
+title: Added support for LFS Download in the importing process
+merge_request: 18871
+author:
+type: fixed

lib/gitlab/git/repository.rb

@@ -1543,7 +1543,7 @@ module Gitlab
         end
       end
-      def rev_list(including: [], excluding: [], objects: false, &block)
+      def rev_list(including: [], excluding: [], options: [], objects: false, &block)
         args = ['rev-list']
         args.push(*rev_list_param(including))
@@ -1556,6 +1556,10 @@ module Gitlab
         args.push('--objects') if objects
+        if options.any?
+          args.push(*options)
+        end
+
         run_git!(args, lazy_block: block)
       end

lib/gitlab/git/rev_list.rb

@@ -38,7 +38,10 @@ module Gitlab
       end
       def all_objects(require_path: nil, &lazy_block)
-        get_objects(including: :all, require_path: require_path, &lazy_block)
+        get_objects(including: :all,
+                    options: ["--filter=blob:limit=#{Gitlab::Git::Blob::LFS_POINTER_MAX_SIZE}"],
+                    require_path: require_path,
+                    &lazy_block)
       end
       # This methods returns an array of missed references
@@ -54,8 +57,8 @@ module Gitlab
         repository.rev_list(args).split("\n")
       end
-      def get_objects(including: [], excluding: [], require_path: nil)
-        opts = { including: including, excluding: excluding, objects: true }
+      def get_objects(including: [], excluding: [], options: [], require_path: nil)
+        opts = { including: including, excluding: excluding, options: options, objects: true }
         repository.rev_list(opts) do |lazy_output|
           objects = objects_from_output(lazy_output, require_path: require_path)

lib/gitlab/github_import/importer/lfs_object_importer.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module GithubImport
+    module Importer
+      class LfsObjectImporter
+        attr_reader :lfs_object, :project
+
+        # lfs_object - An instance of `Gitlab::GithubImport::Representation::LfsObject`.
+        # project - An instance of `Project`.
+        def initialize(lfs_object, project, _)
+          @lfs_object = lfs_object
+          @project = project
+        end
+
+        def execute
+          Projects::LfsPointers::LfsDownloadService
+            .new(project)
+            .execute(lfs_object.oid, lfs_object.download_link)
+        end
+      end
+    end
+  end
+end

lib/gitlab/github_import/importer/lfs_objects_importer.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module GithubImport
+    module Importer
+      class LfsObjectsImporter
+        include ParallelScheduling
+
+        def importer_class
+          LfsObjectImporter
+        end
+
+        def representation_class
+          Representation::LfsObject
+        end
+
+        def sidekiq_worker_class
+          ImportLfsObjectWorker
+        end
+
+        def collection_method
+          :lfs_objects
+        end
+
+        def each_object_to_import
+          lfs_objects = Projects::LfsPointers::LfsImportService.new(project).execute
+
+          lfs_objects.each do |object|
+            yield object
+          end
+        rescue StandardError => e
+          Rails.logger.error("The Lfs import process failed. #{e.message}")
+        end
+      end
+    end
+  end
+end

lib/gitlab/github_import/representation/lfs_object.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module GithubImport
+    module Representation
+      class LfsObject
+        include ToHash
+        include ExposeAttribute
+
+        attr_reader :attributes
+
+        expose_attribute :oid, :download_link
+
+        # Builds a lfs_object
+        def self.from_api_response(lfs_object)
+          new({ oid: lfs_object[0], download_link: lfs_object[1] })
+        end
+
+        # Builds a new lfs_object using a Hash that was built from a JSON payload.
+        def self.from_json_hash(raw_hash)
+          new(Representation.symbolize_hash(raw_hash))
+        end
+
+        # attributes - A Hash containing the raw lfs_object details. The keys of this
+        #              Hash must be Symbols.
+        def initialize(attributes)
+          @attributes = attributes
+        end
+      end
+    end
+  end
+end

lib/gitlab/github_import/sequential_importer.rb

@@ -19,7 +19,8 @@ module Gitlab
         Importer::PullRequestsImporter,
         Importer::IssuesImporter,
         Importer::DiffNotesImporter,
-        Importer::NotesImporter
+        Importer::NotesImporter,
+        Importer::LfsObjectsImporter
       ].freeze
       # project - The project to import the data into.