Merge branch 'tc-backfill-hashed-project_repositories' into 'master'

Fill project_repositories for hashed storage projects

Closes #48527

See merge request gitlab-org/gitlab-ce!23482

changelogs/unreleased/tc-backfill-hashed-project_repositories.yml

@@ -0,0 +1,5 @@
+---
+title: Fill project_repositories for hashed storage projects
+merge_request: 23482
+author:
+type: added

db/post_migrate/20181130102132_backfill_hashed_project_repositories.rb

@@ -0,0 +1,26 @@
+# frozen_string_literal: true
+
+class BackfillHashedProjectRepositories < ActiveRecord::Migration[4.2]
+  include Gitlab::Database::MigrationHelpers
+
+  DOWNTIME = false
+  BATCH_SIZE     = 1_000
+  DELAY_INTERVAL = 5.minutes
+  MIGRATION      = 'BackfillHashedProjectRepositories'
+
+  disable_ddl_transaction!
+
+  class Project < ActiveRecord::Base
+    include EachBatch
+
+    self.table_name = 'projects'
+  end
+
+  def up
+    queue_background_migration_jobs_by_range_at_intervals(Project, MIGRATION, DELAY_INTERVAL)
+  end
+
+  def down
+    # no-op: since there could have been existing rows before the migration do not remove anything
+  end
+end

lib/gitlab/background_migration/backfill_hashed_project_repositories.rb

@@ -0,0 +1,134 @@
+# frozen_string_literal: true
+
+module Gitlab
+  module BackgroundMigration
+    # Class that will create fill the project_repositories table
+    # for all projects that are on hashed storage and an entry is
+    # is missing in this table.
+    class BackfillHashedProjectRepositories
+      # Shard model
+      class Shard < ActiveRecord::Base
+        self.table_name = 'shards'
+      end
+
+      # Class that will find or create the shard by name.
+      # There is only a small set of shards, which would
+      # not change quickly, so look them up from memory
+      # instead of hitting the DB each time.
+      class ShardFinder
+        def find_shard_id(name)
+          shard_id = shards.fetch(name, nil)
+          return shard_id if shard_id.present?
+
+          Shard.transaction(requires_new: true) do
+            create!(name)
+          end
+        rescue ActiveRecord::RecordNotUnique
+          reload!
+          retry
+        end
+
+        private
+
+        def create!(name)
+          Shard.create!(name: name).tap { |shard| @shards[name] = shard.id }
+        end
+
+        def shards
+          @shards ||= reload!
+        end
+
+        def reload!
+          @shards = Hash[*Shard.all.map { |shard| [shard.name, shard.id] }.flatten]
+        end
+      end
+
+      # ProjectRegistry model
+      class ProjectRepository < ActiveRecord::Base
+        self.table_name = 'project_repositories'
+
+        belongs_to :project, inverse_of: :project_repository
+      end
+
+      # Project model
+      class Project < ActiveRecord::Base
+        self.table_name = 'projects'
+
+        HASHED_PATH_PREFIX = '@hashed'
+
+        HASHED_STORAGE_FEATURES = {
+          repository: 1,
+          attachments: 2
+        }.freeze
+
+        has_one :project_repository, inverse_of: :project
+
+        class << self
+          def on_hashed_storage
+            where(Project.arel_table[:storage_version]
+              .gteq(HASHED_STORAGE_FEATURES[:repository]))
+          end
+
+          def without_project_repository
+            joins(left_outer_join_project_repository)
+              .where(ProjectRepository.arel_table[:project_id].eq(nil))
+          end
+
+          def left_outer_join_project_repository
+            projects_table = Project.arel_table
+            repository_table = ProjectRepository.arel_table
+
+            projects_table
+              .join(repository_table, Arel::Nodes::OuterJoin)
+              .on(projects_table[:id].eq(repository_table[:project_id]))
+              .join_sources
+          end
+        end
+
+        def hashed_storage?
+          self.storage_version && self.storage_version >= 1
+        end
+
+        def hashed_disk_path
+          "#{HASHED_PATH_PREFIX}/#{disk_hash[0..1]}/#{disk_hash[2..3]}/#{disk_hash}"
+        end
+
+        def disk_hash
+          @disk_hash ||= Digest::SHA2.hexdigest(id.to_s)
+        end
+      end
+
+      def perform(start_id, stop_id)
+        Gitlab::Database.bulk_insert(:project_repositories, project_repositories(start_id, stop_id))
+      end
+
+      private
+
+      def project_repositories(start_id, stop_id)
+        Project.on_hashed_storage
+          .without_project_repository
+          .where(id: start_id..stop_id)
+          .map { |project| build_attributes_for_project(project) }
+          .compact
+      end
+
+      def build_attributes_for_project(project)
+        return unless project.hashed_storage?
+
+        {
+          project_id: project.id,
+          shard_id:   find_shard_id(project.repository_storage),
+          disk_path:  project.hashed_disk_path
+        }
+      end
+
+      def find_shard_id(repository_storage)
+        shard_finder.find_shard_id(repository_storage)
+      end
+
+      def shard_finder
+        @shard_finder ||= ShardFinder.new
+      end
+    end
+  end
+end

spec/lib/gitlab/background_migration/backfill_hashed_project_repositories_spec.rb

@@ -0,0 +1,90 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+
+describe Gitlab::BackgroundMigration::BackfillHashedProjectRepositories, :migration, schema: 20181130102132 do
+  let(:namespaces) { table(:namespaces) }
+  let(:project_repositories) { table(:project_repositories) }
+  let(:projects) { table(:projects) }
+  let(:shards) { table(:shards) }
+  let(:group) { namespaces.create!(name: 'foo', path: 'foo') }
+  let(:shard) { shards.create!(name: 'default') }
+
+  describe described_class::ShardFinder do
+    describe '#find_shard_id' do
+      it 'creates a new shard when it does not exist yet' do
+        expect { subject.find_shard_id('other') }.to change(shards, :count).by(1)
+      end
+
+      it 'returns the shard when it exists' do
+        shards.create(id: 5, name: 'other')
+
+        shard_id = subject.find_shard_id('other')
+
+        expect(shard_id).to eq(5)
+      end
+
+      it 'only queries the database once to retrieve shards' do
+        subject.find_shard_id('default')
+
+        expect { subject.find_shard_id('default') }.not_to exceed_query_limit(0)
+      end
+    end
+  end
+
+  describe described_class::Project do
+    describe '.on_hashed_storage' do
+      it 'finds projects with repository on hashed storage' do
+        projects.create!(id: 1, name: 'foo', path: 'foo', namespace_id: group.id, storage_version: 1)
+        projects.create!(id: 2, name: 'bar', path: 'bar', namespace_id: group.id, storage_version: 2)
+        projects.create!(id: 3, name: 'baz', path: 'baz', namespace_id: group.id, storage_version: 0)
+        projects.create!(id: 4, name: 'zoo', path: 'zoo', namespace_id: group.id, storage_version: nil)
+
+        expect(described_class.on_hashed_storage.pluck(:id)).to match_array([1, 2])
+      end
+    end
+
+    describe '.without_project_repository' do
+      it 'finds projects which do not have a projects_repositories entry' do
+        projects.create!(id: 1, name: 'foo', path: 'foo', namespace_id: group.id)
+        projects.create!(id: 2, name: 'bar', path: 'bar', namespace_id: group.id)
+        project_repositories.create!(project_id: 2, disk_path: '@phony/foo/bar', shard_id: shard.id)
+
+        expect(described_class.without_project_repository.pluck(:id)).to contain_exactly(1)
+      end
+    end
+  end
+
+  describe '#perform' do
+    it 'creates a project_repository row for projects on hashed storage that need one' do
+      projects.create!(id: 1, name: 'foo', path: 'foo', namespace_id: group.id, storage_version: 1)
+      projects.create!(id: 2, name: 'bar', path: 'bar', namespace_id: group.id, storage_version: 2)
+
+      expect { described_class.new.perform(1, projects.last.id) }.to change(project_repositories, :count).by(2)
+    end
+
+    it 'does nothing for projects on hashed storage that have already a project_repository row' do
+      projects.create!(id: 1, name: 'foo', path: 'foo', namespace_id: group.id, storage_version: 1)
+      project_repositories.create!(project_id: 1, disk_path: '@phony/foo/bar', shard_id: shard.id)
+
+      expect { described_class.new.perform(1, projects.last.id) }.not_to change(project_repositories, :count)
+    end
+
+    it 'does nothing for projects on legacy storage' do
+      projects.create!(name: 'foo', path: 'foo', namespace_id: group.id, storage_version: 0)
+
+      expect { described_class.new.perform(1, projects.last.id) }.not_to change(project_repositories, :count)
+    end
+
+    it 'inserts rows in a single query' do
+      projects.create!(name: 'foo', path: 'foo', namespace_id: group.id, storage_version: 1, repository_storage: shard.name)
+
+      control_count = ActiveRecord::QueryRecorder.new { described_class.new.perform(1, projects.last.id) }
+
+      projects.create!(name: 'bar', path: 'bar', namespace_id: group.id, storage_version: 1, repository_storage: shard.name)
+      projects.create!(name: 'zoo', path: 'zoo', namespace_id: group.id, storage_version: 1, repository_storage: shard.name)
+
+      expect { described_class.new.perform(1, projects.last.id) }.not_to exceed_query_limit(control_count)
+    end
+  end
+end