Merge branch 'fix-github-importer-slowness' into 'master'

app/models/concerns/importable.rb

@@ -3,4 +3,7 @@ module Importable
 
   attr_accessor :importing
   alias_method :importing?, :importing
+
+  attr_accessor :imported
+  alias_method :imported?, :imported
 end

app/models/concerns/issuable.rb

@@ -14,6 +14,7 @@ module Issuable
   include Awardable
   include Taskable
   include TimeTrackable
+  include Importable
 
   # This object is used to gather issuable meta data for displaying
   # upvotes, downvotes, notes and closing merge requests count for issues and merge requests
@@ -99,7 +100,7 @@ module Issuable
     acts_as_paranoid
 
     after_save :update_assignee_cache_counts, if: :assignee_id_changed?
-    after_save :record_metrics
+    after_save :record_metrics, unless: :imported?
 
     def update_assignee_cache_counts
       # make sure we flush the cache for both the old *and* new assignees(if they exist)

app/models/concerns/repository_mirroring.rb

+module RepositoryMirroring
+  def set_remote_as_mirror(name)
+    config = raw_repository.rugged.config
+
+    # This is used to define repository as equivalent as "git clone --mirror"
+    config["remote.#{name}.fetch"] = 'refs/*:refs/*'
+    config["remote.#{name}.mirror"] = true
+    config["remote.#{name}.prune"] = true
+  end
+
+  def fetch_mirror(remote, url)
+    add_remote(remote, url)
+    set_remote_as_mirror(remote)
+    fetch_remote(remote, forced: true)
+    remove_remote(remote)
+  end
+end

app/models/merge_request.rb

@@ -3,7 +3,6 @@ class MergeRequest < ActiveRecord::Base
   include Issuable
   include Referable
   include Sortable
-  include Importable
 
   belongs_to :target_project, class_name: "Project"
   belongs_to :source_project, class_name: "Project"

app/models/project.rb

@@ -551,6 +551,10 @@ class Project < ActiveRecord::Base
     import_type == 'gitea'
   end
 
+  def github_import?
+    import_type == 'github'
+  end
+
   def check_limit
     unless creator.can_create_project? || namespace.kind == 'group'
       projects_limit = creator.projects_limit

app/models/repository.rb

@@ -2,6 +2,7 @@ require 'securerandom'
 
 class Repository
   include Gitlab::ShellAdapter
+  include RepositoryMirroring
 
   attr_accessor :path_with_namespace, :project
 
@@ -64,7 +65,7 @@ class Repository
   # Return absolute path to repository
   def path_to_repo
     @path_to_repo ||= File.expand_path(
-      File.join(@project.repository_storage_path, path_with_namespace + ".git")
+      File.join(repository_storage_path, path_with_namespace + ".git")
     )
   end
 
@@ -401,10 +402,6 @@ class Repository
     expire_tags_cache
   end
 
-  def before_import
-    expire_content_cache
-  end
-
   # Runs code after the HEAD of a repository is changed.
   def after_change_head
     expire_method_caches(METHOD_CACHES_FOR_FILE_TYPES.keys)
@@ -1033,6 +1030,23 @@ class Repository
     rugged.references.delete(tmp_ref) if tmp_ref
   end
 
+  def add_remote(name, url)
+    raw_repository.remote_add(name, url)
+  rescue Rugged::ConfigError
+    raw_repository.remote_update(name, url: url)
+  end
+
+  def remove_remote(name)
+    raw_repository.remote_delete(name)
+    true
+  rescue Rugged::ConfigError
+    false
+  end
+
+  def fetch_remote(remote, forced: false, no_tags: false)
+    gitlab_shell.fetch_remote(repository_storage_path, path_with_namespace, remote, forced: forced, no_tags: no_tags)
+  end
+
   def fetch_ref(source_path, source_ref, target_ref)
     args = %W(#{Gitlab.config.git.bin_path} fetch --no-tags -f #{source_path} #{source_ref}:#{target_ref})
     Gitlab::Popen.popen(args, path_to_repo)
@@ -1150,4 +1164,8 @@ class Repository
   def repository_event(event, tags = {})
     Gitlab::Metrics.add_event(event, { path: path_with_namespace }.merge(tags))
   end
+
+  def repository_storage_path
+    @project.repository_storage_path
+  end
 end

app/services/projects/import_service.rb

@@ -11,7 +11,7 @@ module Projects
 
       success
     rescue => e
-      error(e.message)
+      error("Error importing repository #{project.import_url} into #{project.path_with_namespace} - #{e.message}")
     end
 
     private
@@ -32,23 +32,40 @@ module Projects
     end
 
     def import_repository
+      raise Error, 'Blocked import URL.' if Gitlab::UrlBlocker.blocked_url?(project.import_url)
+
       begin
-        raise Error, "Blocked import URL." if Gitlab::UrlBlocker.blocked_url?(project.import_url)
-        gitlab_shell.import_repository(project.repository_storage_path, project.path_with_namespace, project.import_url)
-      rescue => e
+        if project.github_import? || project.gitea_import?
+          fetch_repository
+        else
+          clone_repository
+        end
+      rescue Gitlab::Shell::Error => e
         # Expire cache to prevent scenarios such as:
         # 1. First import failed, but the repo was imported successfully, so +exists?+ returns true
         # 2. Retried import, repo is broken or not imported but +exists?+ still returns true
-        project.repository.before_import if project.repository_exists?
+        project.repository.expire_content_cache if project.repository_exists?
 
-        raise Error, "Error importing repository #{project.import_url} into #{project.path_with_namespace} - #{e.message}"
+        raise Error, e.message
       end
     end
 
+    def clone_repository
+      gitlab_shell.import_repository(project.repository_storage_path, project.path_with_namespace, project.import_url)
+    end
+
+    def fetch_repository
+      project.create_repository
+      project.repository.add_remote(project.import_type, project.import_url)
+      project.repository.set_remote_as_mirror(project.import_type)
+      project.repository.fetch_remote(project.import_type, forced: true)
+      project.repository.remove_remote(project.import_type)
+    end
+
     def import_data
       return unless has_importer?
 
-      project.repository.before_import unless project.gitlab_project_import?
+      project.repository.expire_content_cache unless project.gitlab_project_import?
 
       unless importer.execute
         raise Error, 'The remote data could not be imported.'

app/workers/repository_import_worker.rb

 class RepositoryImportWorker
   include Sidekiq::Worker
-  include Gitlab::ShellAdapter
   include DedicatedSidekiqQueue
 
   attr_accessor :project, :current_user

changelogs/unreleased/fix-github-importer-slowness.yml

+---
+title: Improve performance of GitHub importer for large repositories.
+merge_request: 10273
+author:

lib/gitlab/github_import/importer.rb

@@ -157,7 +157,7 @@ module Gitlab
       end
 
       def restore_source_branch(pull_request)
-        project.repository.fetch_ref(repo_url, "pull/#{pull_request.number}/head", pull_request.source_branch_name)
+        project.repository.create_branch(pull_request.source_branch_name, pull_request.source_branch_sha)
       end
 
       def restore_target_branch(pull_request)

lib/gitlab/github_import/pull_request_formatter.rb

@@ -20,7 +20,8 @@ module Gitlab
           author_id: author_id,
           assignee_id: assignee_id,
           created_at: raw_data.created_at,
-          updated_at: raw_data.updated_at
+          updated_at: raw_data.updated_at,
+          imported: true
         }
       end
 

lib/gitlab/shell.rb

@@ -88,6 +88,26 @@ module Gitlab
       true
     end
 
+    # Fetch remote for repository
+    #
+    # name - project path with namespace
+    # remote - remote name
+    # forced - should we use --force flag?
+    # no_tags - should we use --no-tags flag?
+    #
+    # Ex.
+    #   fetch_remote("gitlab/gitlab-ci", "upstream")
+    #
+    def fetch_remote(storage, name, remote, forced: false, no_tags: false)
+      args = [gitlab_shell_projects_path, 'fetch-remote', storage, "#{name}.git", remote, '800']
+      args << '--force' if forced
+      args << '--no-tags' if no_tags
+
+      output, status = Popen.popen(args)
+      raise Error, output unless status.zero?
+      true
+    end
+
     # Move repository
     # storage - project's storage path
     # path - project path with namespace

spec/lib/gitlab/github_import/pull_request_formatter_spec.rb

@@ -64,7 +64,8 @@ describe Gitlab::GithubImport::PullRequestFormatter, lib: true do
           author_id: project.creator_id,
           assignee_id: nil,
           created_at: created_at,
-          updated_at: updated_at
+          updated_at: updated_at,
+          imported: true
         }
 
         expect(pull_request.attributes).to eq(expected)
@@ -90,7 +91,8 @@ describe Gitlab::GithubImport::PullRequestFormatter, lib: true do
           author_id: project.creator_id,
           assignee_id: nil,
           created_at: created_at,
-          updated_at: updated_at
+          updated_at: updated_at,
+          imported: true
         }
 
         expect(pull_request.attributes).to eq(expected)
@@ -117,7 +119,8 @@ describe Gitlab::GithubImport::PullRequestFormatter, lib: true do
           author_id: project.creator_id,
           assignee_id: nil,
           created_at: created_at,
-          updated_at: updated_at
+          updated_at: updated_at,
+          imported: true
         }
 
         expect(pull_request.attributes).to eq(expected)

spec/models/repository_spec.rb

@@ -1293,14 +1293,6 @@ describe Repository, models: true do
     end
   end
 
-  describe '#before_import' do
-    it 'flushes the repository caches' do
-      expect(repository).to receive(:expire_content_cache)
-
-      repository.before_import
-    end
-  end
-
   describe '#after_import' do
     it 'flushes and builds the cache' do
       expect(repository).to receive(:expire_content_cache)

spec/services/projects/import_service_spec.rb

@@ -26,30 +26,59 @@ describe Projects::ImportService, services: true do
         result = subject.execute
 
         expect(result[:status]).to eq :error
-        expect(result[:message]).to eq 'The repository could not be created.'
+        expect(result[:message]).to eq "Error importing repository #{project.import_url} into #{project.path_with_namespace} - The repository could not be created."
       end
     end
 
     context 'with known url' do
       before do
         project.import_url = 'https://github.com/vim/vim.git'
+        project.import_type = 'github'
       end
 
-      it 'succeeds if repository import is successfully' do
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).with(project.repository_storage_path, project.path_with_namespace, project.import_url).and_return(true)
+      context 'with a Github repository' do
+        it 'succeeds if repository import is successfully' do
+          expect_any_instance_of(Repository).to receive(:fetch_remote).and_return(true)
+          expect_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_return(true)
 
-        result = subject.execute
+          result = subject.execute
 
-        expect(result[:status]).to eq :success
+          expect(result[:status]).to eq :success
+        end
+
+        it 'fails if repository import fails' do
+          expect_any_instance_of(Repository).to receive(:fetch_remote).and_raise(Gitlab::Shell::Error.new('Failed to import the repository'))
+
+          result = subject.execute
+
+          expect(result[:status]).to eq :error
+          expect(result[:message]).to eq "Error importing repository #{project.import_url} into #{project.path_with_namespace} - Failed to import the repository"
+        end
       end
 
-      it 'fails if repository import fails' do
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).with(project.repository_storage_path, project.path_with_namespace, project.import_url).and_raise(Gitlab::Shell::Error.new('Failed to import the repository'))
+      context 'with a non Github repository' do
+        before do
+          project.import_url = 'https://bitbucket.org/vim/vim.git'
+          project.import_type = 'bitbucket'
+        end
 
-        result = subject.execute
+        it 'succeeds if repository import is successfully' do
+          expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).and_return(true)
+          expect_any_instance_of(Gitlab::BitbucketImport::Importer).to receive(:execute).and_return(true)
 
-        expect(result[:status]).to eq :error
-        expect(result[:message]).to eq "Error importing repository #{project.import_url} into #{project.path_with_namespace} - Failed to import the repository"
+          result = subject.execute
+
+          expect(result[:status]).to eq :success
+        end
+
+        it 'fails if repository import fails' do
+          expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).and_raise(Gitlab::Shell::Error.new('Failed to import the repository'))
+
+          result = subject.execute
+
+          expect(result[:status]).to eq :error
+          expect(result[:message]).to eq "Error importing repository #{project.import_url} into #{project.path_with_namespace} - Failed to import the repository"
+        end
       end
     end
 
@@ -64,8 +93,8 @@ describe Projects::ImportService, services: true do
       end
 
       it 'succeeds if importer succeeds' do
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).with(project.repository_storage_path, project.path_with_namespace, project.import_url).and_return(true)
-        expect_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_return(true)
+        allow_any_instance_of(Repository).to receive(:fetch_remote).and_return(true)
+        allow_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_return(true)
 
         result = subject.execute
 
@@ -73,48 +102,42 @@ describe Projects::ImportService, services: true do
       end
 
       it 'flushes various caches' do
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).
-          with(project.repository_storage_path, project.path_with_namespace, project.import_url).
+        allow_any_instance_of(Repository).to receive(:fetch_remote).
           and_return(true)
 
-        expect_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).
+        allow_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).
           and_return(true)
 
-        expect_any_instance_of(Repository).to receive(:expire_emptiness_caches).
-          and_call_original
-
-        expect_any_instance_of(Repository).to receive(:expire_exists_cache).
-          and_call_original
+        expect_any_instance_of(Repository).to receive(:expire_content_cache)
 
         subject.execute
       end
 
       it 'fails if importer fails' do
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).with(project.repository_storage_path, project.path_with_namespace, project.import_url).and_return(true)
-        expect_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_return(false)
+        allow_any_instance_of(Repository).to receive(:fetch_remote).and_return(true)
+        allow_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_return(false)
 
         result = subject.execute
 
         expect(result[:status]).to eq :error
-        expect(result[:message]).to eq 'The remote data could not be imported.'
+        expect(result[:message]).to eq "Error importing repository #{project.import_url} into #{project.path_with_namespace} - The remote data could not be imported."
       end
 
       it 'fails if importer raise an error' do
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).with(project.repository_storage_path, project.path_with_namespace, project.import_url).and_return(true)
-        expect_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_raise(Projects::ImportService::Error.new('Github: failed to connect API'))
+        allow_any_instance_of(Gitlab::Shell).to receive(:fetch_remote).and_return(true)
+        allow_any_instance_of(Gitlab::GithubImport::Importer).to receive(:execute).and_raise(Projects::ImportService::Error.new('Github: failed to connect API'))
 
         result = subject.execute
 
         expect(result[:status]).to eq :error
-        expect(result[:message]).to eq 'Github: failed to connect API'
+        expect(result[:message]).to eq "Error importing repository #{project.import_url} into #{project.path_with_namespace} - Github: failed to connect API"
       end
 
-      it 'expires existence cache after error' do
+      it 'expires content cache after error' do
         allow_any_instance_of(Project).to receive(:repository_exists?).and_return(false, true)
 
-        expect_any_instance_of(Gitlab::Shell).to receive(:import_repository).with(project.repository_storage_path, project.path_with_namespace, project.import_url).and_raise(Gitlab::Shell::Error.new('Failed to import the repository'))
-        expect_any_instance_of(Repository).to receive(:expire_emptiness_caches).and_call_original
-        expect_any_instance_of(Repository).to receive(:expire_exists_cache).and_call_original
+        expect_any_instance_of(Gitlab::Shell).to receive(:fetch_remote).and_raise(Gitlab::Shell::Error.new('Failed to import the repository'))
+        expect_any_instance_of(Repository).to receive(:expire_content_cache)
 
         subject.execute
       end