Merge branch '52300-pool-repositories' into 'master'

Start tracking pool repositories

Closes #38238 and #52300

See merge request gitlab-org/gitlab-ce!22482

app/models/pool_repository.rb

+# frozen_string_literal: true
+
+class PoolRepository < ActiveRecord::Base
+  POOL_PREFIX = '@pools'
+
+  belongs_to :shard
+  validates :shard, presence: true
+
+  # For now, only pool repositories are tracked in the database. However, we may
+  # want to add other repository types in the future
+  self.table_name = 'repositories'
+
+  has_many :pool_member_projects, class_name: 'Project', foreign_key: :pool_repository_id
+
+  def shard_name
+    shard&.name
+  end
+
+  def shard_name=(name)
+    self.shard = Shard.by_name(name)
+  end
+end

app/models/project.rb

@@ -124,6 +124,7 @@ class Project < ActiveRecord::Base
   alias_attribute :title, :name
 
   # Relations
+  belongs_to :pool_repository
   belongs_to :creator, class_name: 'User'
   belongs_to :group, -> { where(type: 'Group') }, foreign_key: 'namespace_id'
   belongs_to :namespace

app/models/shard.rb

+# frozen_string_literal: true
+
+class Shard < ActiveRecord::Base
+  # Store shard names from the configuration file in the database. This is not a
+  # list of active shards - we just want to assign an immutable, unique ID to
+  # every shard name for easy indexing / referencing.
+  def self.populate!
+    return unless table_exists?
+
+    # The GitLab config does not change for the lifecycle of the process
+    in_config = Gitlab.config.repositories.storages.keys.map(&:to_s)
+
+    transaction do
+      in_db = all.pluck(:name)
+      missing = in_config - in_db
+
+      missing.map { |name| by_name(name) }
+    end
+  end
+
+  def self.by_name(name)
+    find_or_create_by(name: name)
+  rescue ActiveRecord::RecordNotUnique
+    retry
+  end
+end

changelogs/unreleased/52300-pool-repositories.yml

+---
+title: Start tracking shards and pool repositories in the database
+merge_request: 22482
+author:
+type: other

config/initializers/fill_shards.rb

+return unless Shard.connected?
+return if Gitlab::Database.read_only?
+
+Shard.populate!

db/migrate/20181019032400_add_shards_table.rb

+# frozen_string_literal: true
+
+class AddShardsTable < ActiveRecord::Migration
+  DOWNTIME = false
+
+  def change
+    create_table :shards do |t|
+      t.string :name, null: false, index: { unique: true }
+    end
+  end
+end

db/migrate/20181019032408_add_repositories_table.rb

+# frozen_string_literal: true
+
+class AddRepositoriesTable < ActiveRecord::Migration
+  DOWNTIME = false
+
+  def change
+    create_table :repositories, id: :bigserial do |t|
+      t.references :shard, null: false, index: true, foreign_key: { on_delete: :restrict }
+      t.string :disk_path, null: false, index: { unique: true }
+    end
+
+    add_column :projects, :pool_repository_id, :bigint
+    add_index :projects, :pool_repository_id, where: 'pool_repository_id IS NOT NULL'
+  end
+end

db/migrate/20181019105553_add_projects_pool_repository_id_foreign_key.rb

+# frozen_string_literal: true
+
+class AddProjectsPoolRepositoryIdForeignKey < ActiveRecord::Migration
+  include Gitlab::Database::MigrationHelpers
+
+  DOWNTIME = false
+
+  disable_ddl_transaction!
+
+  def up
+    add_concurrent_foreign_key(
+      :projects,
+      :repositories,
+      column: :pool_repository_id,
+      on_delete: :nullify
+    )
+  end
+
+  def down
+    remove_foreign_key(:projects, column: :pool_repository_id)
+  end
+end

db/schema.rb

@@ -1703,6 +1703,7 @@ ActiveRecord::Schema.define(version: 20181101144347) do
     t.integer "jobs_cache_index"
     t.boolean "pages_https_only", default: true
     t.boolean "remote_mirror_available_overridden"
+    t.integer "pool_repository_id", limit: 8
   end
 
   add_index "projects", ["ci_id"], name: "index_projects_on_ci_id", using: :btree
@@ -1719,6 +1720,7 @@ ActiveRecord::Schema.define(version: 20181101144347) do
   add_index "projects", ["path"], name: "index_projects_on_path", using: :btree
   add_index "projects", ["path"], name: "index_projects_on_path_trigram", using: :gin, opclasses: {"path"=>"gin_trgm_ops"}
   add_index "projects", ["pending_delete"], name: "index_projects_on_pending_delete", using: :btree
+  add_index "projects", ["pool_repository_id"], name: "index_projects_on_pool_repository_id", where: "(pool_repository_id IS NOT NULL)", using: :btree
   add_index "projects", ["repository_storage", "created_at"], name: "idx_project_repository_check_partial", where: "(last_repository_check_at IS NULL)", using: :btree
   add_index "projects", ["repository_storage"], name: "index_projects_on_repository_storage", using: :btree
   add_index "projects", ["runners_token"], name: "index_projects_on_runners_token", using: :btree
@@ -1851,6 +1853,14 @@ ActiveRecord::Schema.define(version: 20181101144347) do
   add_index "remote_mirrors", ["last_successful_update_at"], name: "index_remote_mirrors_on_last_successful_update_at", using: :btree
   add_index "remote_mirrors", ["project_id"], name: "index_remote_mirrors_on_project_id", using: :btree
 
+  create_table "repositories", id: :bigserial, force: :cascade do |t|
+    t.integer "shard_id", null: false
+    t.string "disk_path", null: false
+  end
+
+  add_index "repositories", ["disk_path"], name: "index_repositories_on_disk_path", unique: true, using: :btree
+  add_index "repositories", ["shard_id"], name: "index_repositories_on_shard_id", using: :btree
+
   create_table "repository_languages", id: false, force: :cascade do |t|
     t.integer "project_id", null: false
     t.integer "programming_language_id", null: false
@@ -1931,6 +1941,12 @@ ActiveRecord::Schema.define(version: 20181101144347) do
   add_index "services", ["project_id"], name: "index_services_on_project_id", using: :btree
   add_index "services", ["template"], name: "index_services_on_template", using: :btree
 
+  create_table "shards", force: :cascade do |t|
+    t.string "name", null: false
+  end
+
+  add_index "shards", ["name"], name: "index_shards_on_name", unique: true, using: :btree
+
   create_table "site_statistics", force: :cascade do |t|
     t.integer "repositories_count", default: 0, null: false
   end
@@ -2450,6 +2466,7 @@ ActiveRecord::Schema.define(version: 20181101144347) do
   add_foreign_key "project_import_data", "projects", name: "fk_ffb9ee3a10", on_delete: :cascade
   add_foreign_key "project_mirror_data", "projects", on_delete: :cascade
   add_foreign_key "project_statistics", "projects", on_delete: :cascade
+  add_foreign_key "projects", "repositories", column: "pool_repository_id", name: "fk_6e5c14658a", on_delete: :nullify
   add_foreign_key "prometheus_metrics", "projects", on_delete: :cascade
   add_foreign_key "protected_branch_merge_access_levels", "protected_branches", name: "fk_8a3072ccb3", on_delete: :cascade
   add_foreign_key "protected_branch_push_access_levels", "protected_branches", name: "fk_9ffc86a3d9", on_delete: :cascade
@@ -2461,6 +2478,7 @@ ActiveRecord::Schema.define(version: 20181101144347) do
   add_foreign_key "push_event_payloads", "events", name: "fk_36c74129da", on_delete: :cascade
   add_foreign_key "releases", "projects", name: "fk_47fe2a0596", on_delete: :cascade
   add_foreign_key "remote_mirrors", "projects", on_delete: :cascade
+  add_foreign_key "repositories", "shards", on_delete: :restrict
   add_foreign_key "repository_languages", "projects", on_delete: :cascade
   add_foreign_key "resource_label_events", "issues", on_delete: :cascade
   add_foreign_key "resource_label_events", "labels", on_delete: :nullify

lib/gitlab/import_export/import_export.yml

@@ -92,6 +92,7 @@ excluded_attributes:
     - :path
     - :namespace_id
     - :creator_id
+    - :pool_repository_id
     - :import_url
     - :import_status
     - :avatar

spec/lib/gitlab/import_export/all_models.yml

@@ -299,6 +299,7 @@ project:
 - ci_cd_settings
 - import_export_upload
 - repository_languages
+- pool_repository
 award_emoji:
 - awardable
 - user

spec/models/project_spec.rb

@@ -8,6 +8,7 @@ describe Project do
     it { is_expected.to belong_to(:group) }
     it { is_expected.to belong_to(:namespace) }
     it { is_expected.to belong_to(:creator).class_name('User') }
+    it { is_expected.to belong_to(:pool_repository) }
     it { is_expected.to have_many(:users) }
     it { is_expected.to have_many(:services) }
     it { is_expected.to have_many(:events) }

spec/models/shard_spec.rb

+# frozen_string_literals: true
+require 'spec_helper'
+
+describe Shard do
+  describe '.populate!' do
+    it 'creates shards based on the config file' do
+      expect(described_class.all).to be_empty
+
+      stub_storage_settings(foo: {}, bar: {}, baz: {})
+
+      described_class.populate!
+
+      expect(described_class.all.map(&:name)).to match_array(%w[default foo bar baz])
+    end
+  end
+
+  describe '.by_name' do
+    let(:default_shard) { described_class.find_by(name: 'default') }
+
+    before do
+      described_class.populate!
+    end
+
+    it 'returns an existing shard' do
+      expect(described_class.by_name('default')).to eq(default_shard)
+    end
+
+    it 'creates a new shard' do
+      result = described_class.by_name('foo')
+
+      expect(result).not_to eq(default_shard)
+      expect(result.name).to eq('foo')
+    end
+
+    it 'retries if creation races' do
+      expect(described_class)
+        .to receive(:find_or_create_by)
+        .with(name: 'default')
+        .and_raise(ActiveRecord::RecordNotUnique, 'fail')
+        .once
+
+      expect(described_class)
+        .to receive(:find_or_create_by)
+        .with(name: 'default')
+        .and_call_original
+
+      expect(described_class.by_name('default')).to eq(default_shard)
+    end
+  end
+end