Add a migration to populate fork networks

This uses the existing ForkedProjectLinks

db/migrate/20170929131201_populate_fork_networks.rb

+# See http://doc.gitlab.com/ce/development/migration_style_guide.html
+# for more information on how to write migrations for GitLab.
+
+class PopulateForkNetworks < ActiveRecord::Migration
+  include Gitlab::Database::MigrationHelpers
+
+  DOWNTIME = false
+
+  MIGRATION = 'PopulateForkNetworksRange'.freeze
+  BATCH_SIZE = 100
+  DELAY_INTERVAL = 15.seconds
+
+  disable_ddl_transaction!
+
+  class ForkedProjectLink < ActiveRecord::Base
+    include EachBatch
+
+    self.table_name = 'forked_project_links'
+  end
+
+  def up
+    say 'Populating the `fork_networks` based on existing `forked_project_links`'
+
+    queue_background_migration_jobs_by_range_at_intervals(ForkedProjectLink, MIGRATION, DELAY_INTERVAL, batch_size: BATCH_SIZE)
+  end
+
+  def down
+    # nothing
+  end
+end

lib/gitlab/background_migration/create_fork_network_memberships_range.rb

+module Gitlab
+  module BackgroundMigration
+    class CreateForkNetworkMembershipsRange
+      RESCHEDULE_DELAY = 15
+
+      class ForkedProjectLink < ActiveRecord::Base
+        self.table_name = 'forked_project_links'
+      end
+
+      def perform(start_id, end_id)
+        log("Creating memberships for forks: #{start_id} - #{end_id}")
+
+        ActiveRecord::Base.connection.execute <<~INSERT_MEMBERS
+          INSERT INTO fork_network_members (fork_network_id, project_id, forked_from_project_id)
+
+          SELECT fork_network_members.fork_network_id,
+                 forked_project_links.forked_to_project_id,
+                 forked_project_links.forked_from_project_id
+
+          FROM forked_project_links
+
+          INNER JOIN fork_network_members
+              ON forked_project_links.forked_from_project_id = fork_network_members.project_id
+
+          WHERE forked_project_links.id BETWEEN #{start_id} AND #{end_id}
+          AND NOT EXISTS (
+            SELECT true
+            FROM fork_network_members existing_members
+            WHERE existing_members.project_id = forked_project_links.forked_to_project_id
+          )
+        INSERT_MEMBERS
+
+        if missing_members?(start_id, end_id)
+          BackgroundMigrationWorker.perform_in(RESCHEDULE_DELAY, "CreateForkNetworkMembershipsRange", [start_id, end_id])
+        end
+      end
+
+      def missing_members?(start_id, end_id)
+        count_sql = <<~MISSING_MEMBERS
+          SELECT COUNT(*)
+
+          FROM forked_project_links
+
+          WHERE NOT EXISTS (
+            SELECT true
+            FROM fork_network_members
+            WHERE fork_network_members.project_id = forked_project_links.forked_to_project_id
+          )
+          AND forked_project_links.id BETWEEN #{start_id} AND #{end_id}
+        MISSING_MEMBERS
+
+        ForkNetworkMember.count_by_sql(count_sql) > 0
+      end
+
+      def log(message)
+        Rails.logger.info("#{self.class.name} - #{message}")
+      end
+    end
+  end
+end

lib/gitlab/background_migration/populate_fork_networks_range.rb

+module Gitlab
+  module BackgroundMigration
+    class PopulateForkNetworksRange
+      def perform(start_id, end_id)
+        log("Creating fork networks for forked project links: #{start_id} - #{end_id}")
+
+        ActiveRecord::Base.connection.execute <<~INSERT_NETWORKS
+        INSERT INTO fork_networks (root_project_id)
+          SELECT DISTINCT forked_project_links.forked_from_project_id
+
+          FROM forked_project_links
+
+          WHERE NOT EXISTS (
+            SELECT true
+            FROM forked_project_links inner_links
+            WHERE inner_links.forked_to_project_id = forked_project_links.forked_from_project_id
+          )
+          AND NOT EXISTS (
+            SELECT true
+            FROM fork_networks
+            WHERE forked_project_links.forked_from_project_id = fork_networks.root_project_id
+          )
+          AND forked_project_links.id BETWEEN #{start_id} AND #{end_id}
+        INSERT_NETWORKS
+
+        log("Creating memberships for root projects: #{start_id} - #{end_id}")
+
+        ActiveRecord::Base.connection.execute <<~INSERT_ROOT
+          INSERT INTO fork_network_members (fork_network_id, project_id)
+          SELECT DISTINCT fork_networks.id, fork_networks.root_project_id
+
+          FROM fork_networks
+
+          INNER JOIN forked_project_links
+              ON forked_project_links.forked_from_project_id = fork_networks.root_project_id
+
+          WHERE NOT EXISTS (
+            SELECT true
+            FROM fork_network_members
+            WHERE fork_network_members.project_id = fork_networks.root_project_id
+          )
+          AND forked_project_links.id BETWEEN #{start_id} AND #{end_id}
+        INSERT_ROOT
+
+        delay = BackgroundMigration::CreateForkNetworkMembershipsRange::RESCHEDULE_DELAY
+        BackgroundMigrationWorker.perform_in(delay, "CreateForkNetworkMembershipsRange", [start_id, end_id])
+      end
+
+      def log(message)
+        Rails.logger.info("#{self.class.name} - #{message}")
+      end
+    end
+  end
+end

spec/lib/gitlab/background_migration/create_fork_network_memberships_range_spec.rb

+require 'spec_helper'
+
+describe Gitlab::BackgroundMigration::CreateForkNetworkMembershipsRange, :migration, schema: 20170929131201 do
+  let(:migration) { described_class.new }
+
+  let(:base1) { create(:project) }
+  let(:base1_fork1) { create(:project) }
+  let(:base1_fork2) { create(:project) }
+
+  let(:base2) { create(:project) }
+  let(:base2_fork1) { create(:project) }
+  let(:base2_fork2) { create(:project) }
+
+  let(:fork_of_fork) { create(:project) }
+  let(:fork_of_fork2) { create(:project) }
+  let(:second_level_fork) { create(:project) }
+  let(:third_level_fork) { create(:project) }
+
+  let(:fork_network1) { fork_networks.find_by(root_project_id: base1.id) }
+  let(:fork_network2) { fork_networks.find_by(root_project_id: base2.id) }
+
+  let!(:forked_project_links) { table(:forked_project_links) }
+  let!(:fork_networks) { table(:fork_networks) }
+  let!(:fork_network_members) { table(:fork_network_members) }
+
+  before do
+    # The fork-network relation created for the forked project
+    fork_networks.create(id: 1, root_project_id: base1.id)
+    fork_network_members.create(project_id: base1.id, fork_network_id: 1)
+    fork_networks.create(id: 2, root_project_id: base2.id)
+    fork_network_members.create(project_id: base2.id, fork_network_id: 2)
+
+    # Normal fork links
+    forked_project_links.create(id: 1, forked_from_project_id: base1.id, forked_to_project_id: base1_fork1.id)
+    forked_project_links.create(id: 2, forked_from_project_id: base1.id, forked_to_project_id: base1_fork2.id)
+    forked_project_links.create(id: 3, forked_from_project_id: base2.id, forked_to_project_id: base2_fork1.id)
+    forked_project_links.create(id: 4, forked_from_project_id: base2.id, forked_to_project_id: base2_fork2.id)
+
+    # Fork links
+    forked_project_links.create(id: 5, forked_from_project_id: base1_fork1.id, forked_to_project_id: fork_of_fork.id)
+    forked_project_links.create(id: 6, forked_from_project_id: base1_fork1.id, forked_to_project_id: fork_of_fork2.id)
+
+    # Forks 3 levels down
+    forked_project_links.create(id: 7, forked_from_project_id: fork_of_fork.id, forked_to_project_id: second_level_fork.id)
+    forked_project_links.create(id: 8, forked_from_project_id: second_level_fork.id, forked_to_project_id: third_level_fork.id)
+
+    migration.perform(1, 8)
+  end
+
+  it 'creates a memberships for the direct forks' do
+    base1_fork1_membership = fork_network_members.find_by(fork_network_id: fork_network1.id,
+                                                          project_id: base1_fork1.id)
+    base1_fork2_membership = fork_network_members.find_by(fork_network_id: fork_network1.id,
+                                                          project_id: base1_fork2.id)
+    base2_fork1_membership = fork_network_members.find_by(fork_network_id: fork_network2.id,
+                                                          project_id: base2_fork1.id)
+    base2_fork2_membership = fork_network_members.find_by(fork_network_id: fork_network2.id,
+                                                          project_id: base2_fork2.id)
+
+    expect(base1_fork1_membership.forked_from_project_id).to eq(base1.id)
+    expect(base1_fork2_membership.forked_from_project_id).to eq(base1.id)
+    expect(base2_fork1_membership.forked_from_project_id).to eq(base2.id)
+    expect(base2_fork2_membership.forked_from_project_id).to eq(base2.id)
+  end
+
+  it 'adds the fork network members for forks of forks' do
+    fork_of_fork_membership = fork_network_members.find_by(project_id: fork_of_fork.id,
+                                                           fork_network_id: fork_network1.id)
+    fork_of_fork2_membership = fork_network_members.find_by(project_id: fork_of_fork2.id,
+                                                            fork_network_id: fork_network1.id)
+    second_level_fork_membership = fork_network_members.find_by(project_id: second_level_fork.id,
+                                                                fork_network_id: fork_network1.id)
+    third_level_fork_membership = fork_network_members.find_by(project_id: third_level_fork.id,
+                                                               fork_network_id: fork_network1.id)
+
+    expect(fork_of_fork_membership.forked_from_project_id).to eq(base1_fork1.id)
+    expect(fork_of_fork2_membership.forked_from_project_id).to eq(base1_fork1.id)
+    expect(second_level_fork_membership.forked_from_project_id).to eq(fork_of_fork.id)
+    expect(third_level_fork_membership.forked_from_project_id).to eq(second_level_fork.id)
+  end
+
+  it 'reschedules itself when there are missing members' do
+    allow(migration).to receive(:missing_members?).and_return(true)
+
+    expect(BackgroundMigrationWorker)
+      .to receive(:perform_in).with(described_class::RESCHEDULE_DELAY, "CreateForkNetworkMembershipsRange", [1, 3])
+
+    migration.perform(1, 3)
+  end
+
+  it 'can be repeated without effect' do
+    expect { fork_network_members.count }.not_to change { migration.perform(1, 7) }
+  end
+
+  it 'knows it is finished for this range' do
+    expect(migration.missing_members?(1, 7)).to be_falsy
+  end
+
+  context 'with more forks' do
+    before do
+      forked_project_links.create(id: 9, forked_from_project_id: fork_of_fork.id, forked_to_project_id: create(:project).id)
+      forked_project_links.create(id: 10, forked_from_project_id: fork_of_fork.id, forked_to_project_id: create(:project).id)
+    end
+
+    it 'only processes a single batch of links at a time' do
+      expect(fork_network_members.count).to eq(10)
+
+      migration.perform(8, 10)
+
+      expect(fork_network_members.count).to eq(12)
+    end
+
+    it 'knows when not all memberships withing a batch have been created' do
+      expect(migration.missing_members?(8, 10)).to be_truthy
+    end
+  end
+end

spec/lib/gitlab/background_migration/populate_fork_networks_range_spec.rb

+require 'spec_helper'
+
+describe Gitlab::BackgroundMigration::PopulateForkNetworksRange, :migration, schema: 20170929131201 do
+  let(:migration) { described_class.new }
+  let(:base1) { create(:project) }
+  let(:base1_fork1) { create(:project) }
+  let(:base1_fork2) { create(:project) }
+
+  let(:base2) { create(:project) }
+  let(:base2_fork1) { create(:project) }
+  let(:base2_fork2) { create(:project) }
+
+  let!(:forked_project_links) { table(:forked_project_links) }
+  let!(:fork_networks) { table(:fork_networks) }
+  let!(:fork_network_members) { table(:fork_network_members) }
+
+  let(:fork_network1) { fork_networks.find_by(root_project_id: base1.id) }
+  let(:fork_network2) { fork_networks.find_by(root_project_id: base2.id) }
+
+  before do
+    # A normal fork link
+    forked_project_links.create(id: 1,
+                                forked_from_project_id: base1.id,
+                                forked_to_project_id: base1_fork1.id)
+    forked_project_links.create(id: 2,
+                                forked_from_project_id: base1.id,
+                                forked_to_project_id: base1_fork2.id)
+
+    forked_project_links.create(id: 3,
+                                forked_from_project_id: base2.id,
+                                forked_to_project_id: base2_fork1.id)
+    forked_project_links.create(id: 4,
+                                forked_from_project_id: base2_fork1.id,
+                                forked_to_project_id: create(:project).id)
+
+    forked_project_links.create(id: 5,
+                                forked_from_project_id: base2.id,
+                                forked_to_project_id: base2_fork2.id)
+
+    migration.perform(1, 3)
+  end
+
+  it 'it creates the fork network' do
+    expect(fork_network1).not_to be_nil
+    expect(fork_network2).not_to be_nil
+  end
+
+  it 'does not create a fork network for a fork-of-fork' do
+    # perfrom the entire batch
+    migration.perform(1, 5)
+
+    expect(fork_networks.find_by(root_project_id: base2_fork1.id)).to be_nil
+  end
+
+  it 'creates memberships for the root of fork networks' do
+    base1_membership = fork_network_members.find_by(fork_network_id: fork_network1.id,
+                                                    project_id: base1.id)
+    base2_membership = fork_network_members.find_by(fork_network_id: fork_network2.id,
+                                                    project_id: base2.id)
+
+    expect(base1_membership).not_to be_nil
+    expect(base2_membership).not_to be_nil
+  end
+
+  it 'schedules a job for inserting memberships for forks-of-forks' do
+    delay = Gitlab::BackgroundMigration::CreateForkNetworkMembershipsRange::RESCHEDULE_DELAY
+
+    expect(BackgroundMigrationWorker)
+      .to receive(:perform_in).with(delay, "CreateForkNetworkMembershipsRange", [1, 3])
+
+    migration.perform(1, 3)
+  end
+
+  it 'only processes a single batch of links at a time' do
+    expect(fork_network_members.count).to eq(5)
+
+    migration.perform(3, 5)
+
+    expect(fork_network_members.count).to eq(7)
+  end
+
+  it 'can be repeated without effect' do
+    expect { migration.perform(1, 3) }.not_to change { fork_network_members.count }
+  end
+end