Merge branch 'migration-helpers' into 'master'

Added helper methods for database migrations

These helpers can ultimately be used to write migrations that don't
require downtime.

See #15464 for more information.

See merge request !3860

.rubocop.yml

@@ -21,6 +21,7 @@ AllCops:
     - 'lib/email_validator.rb'
     - 'lib/gitlab/upgrader.rb'
     - 'lib/gitlab/seeder.rb'
+    - 'lib/templates/**/*'
 
 
 ##################### Style ##################################

doc/development/migration_style_guide.md

@@ -8,7 +8,10 @@ In addition, having to take a server offline for a an upgrade small or big is
 a big burden for most organizations. For this reason it is important that your
 migrations are written carefully, can be applied online and adhere to the style guide below.
 
-It's advised to have offline migrations only in major GitLab releases.
+Migrations should not require GitLab installations to be taken offline unless
+_absolutely_ necessary. If a migration requires downtime this should be
+clearly mentioned during the review process as well as being documented in the
+monthly release post.
 
 When writing your migrations, also consider that databases might have stale data
 or inconsistencies and guard for that. Try to make as little assumptions as possible
@@ -58,6 +61,45 @@ remove_index :namespaces, column: :name if index_exists?(:namespaces, :name)
 
 If you need to add an unique index please keep in mind there is possibility of existing duplicates. If it is possible write a separate migration for handling this situation. It can be just removing or removing with overwriting all references to these duplicates depend on situation.
 
+When adding an index make sure to use the method `add_concurrent_index` instead
+of the regular `add_index` method. The `add_concurrent_index` method
+automatically creates concurrent indexes when using PostgreSQL, removing the
+need for downtime. To use this method you must disable transactions by calling
+the method `disable_ddl_transaction!` in the body of your migration class like
+so:
+
+```
+class MyMigration < ActiveRecord::Migration
+  disable_ddl_transaction!
+
+  def change
+
+  end
+end
+```
+
+## Adding Columns With Default Values
+
+When adding columns with default values you should use the method
+`add_column_with_default`. This method ensures the table is updated without
+requiring downtime. This method is not reversible so you must manually define
+the `up` and `down` methods in your migration class.
+
+For example, to add the column `foo` to the `projects` table with a default
+value of `10` you'd write the following:
+
+```
+class MyMigration < ActiveRecord::Migration
+  def up
+    add_column_with_default(:projects, :foo, :integer, 10)
+  end
+
+  def down
+    remove_column(:projects, :foo)
+  end
+end
+```
+
 ## Testing
 
 Make sure that your migration works with MySQL and PostgreSQL with data. An empty database does not guarantee that your migration is correct.

lib/gitlab/database/migration_helpers.rb

+module Gitlab
+  module Database
+    module MigrationHelpers
+      # Creates a new index, concurrently when supported
+      #
+      # On PostgreSQL this method creates an index concurrently, on MySQL this
+      # creates a regular index.
+      #
+      # Example:
+      #
+      #     add_concurrent_index :users, :some_column
+      #
+      # See Rails' `add_index` for more info on the available arguments.
+      def add_concurrent_index(*args)
+        if transaction_open?
+          raise 'add_concurrent_index can not be run inside a transaction, ' \
+            'you can disable transactions by calling disable_ddl_transaction! ' \
+            'in the body of your migration class'
+        end
+
+        if Database.postgresql?
+          args << { algorithm: :concurrently }
+        end
+
+        add_index(*args)
+      end
+
+      # Updates the value of a column in batches.
+      #
+      # This method updates the table in batches of 5% of the total row count.
+      # Any data inserted while running this method (or after it has finished
+      # running) is _not_ updated automatically.
+      #
+      # This method _only_ updates rows where the column's value is set to NULL.
+      #
+      # table - The name of the table.
+      # column - The name of the column to update.
+      # value - The value for the column.
+      def update_column_in_batches(table, column, value)
+        quoted_table = quote_table_name(table)
+        quoted_column = quote_column_name(column)
+        quoted_value = quote(value)
+        processed = 0
+
+        total = exec_query("SELECT COUNT(*) AS count FROM #{quoted_table}").
+          to_hash.
+          first['count'].
+          to_i
+
+        # Update in batches of 5% with an upper limit of 5000 rows.
+        batch_size = ((total / 100.0) * 5.0).ceil
+
+        while processed < total
+          start_row = exec_query(%Q{
+            SELECT id
+            FROM #{quoted_table}
+            ORDER BY id ASC
+            LIMIT 1 OFFSET #{processed}
+          }).to_hash.first
+
+          stop_row = exec_query(%Q{
+            SELECT id
+            FROM #{quoted_table}
+            ORDER BY id ASC
+            LIMIT 1 OFFSET #{processed + batch_size}
+          }).to_hash.first
+
+          query = %Q{
+            UPDATE #{quoted_table}
+            SET #{quoted_column} = #{quoted_value}
+            WHERE id >= #{start_row['id']}
+          }
+
+          if stop_row
+            query += " AND id < #{stop_row['id']}"
+          end
+
+          execute(query)
+
+          processed += batch_size
+        end
+      end
+
+      # Adds a column with a default value without locking an entire table.
+      #
+      # This method runs the following steps:
+      #
+      # 1. Add the column with a default value of NULL.
+      # 2. Update all existing rows in batches.
+      # 3. Change the default value of the column to the specified value.
+      # 4. Update any remaining rows.
+      #
+      # These steps ensure a column can be added to a large and commonly used
+      # table without locking the entire table for the duration of the table
+      # modification.
+      #
+      # table - The name of the table to update.
+      # column - The name of the column to add.
+      # type - The column type (e.g. `:integer`).
+      # default - The default value for the column.
+      # allow_null - When set to `true` the column will allow NULL values, the
+      #              default is to not allow NULL values.
+      def add_column_with_default(table, column, type, default:, allow_null: false)
+        if transaction_open?
+          raise 'add_column_with_default can not be run inside a transaction, ' \
+            'you can disable transactions by calling disable_ddl_transaction! ' \
+            'in the body of your migration class'
+        end
+
+        transaction do
+          add_column(table, column, type, default: nil)
+
+          # Changing the default before the update ensures any newly inserted
+          # rows already use the proper default value.
+          change_column_default(table, column, default)
+        end
+
+        begin
+          transaction do
+            update_column_in_batches(table, column, default)
+          end
+        # We want to rescue _all_ exceptions here, even those that don't inherit
+        # from StandardError.
+        rescue Exception => error # rubocop: disable all
+          remove_column(table, column)
+
+          raise error
+        end
+
+        change_column_null(table, column, false) unless allow_null
+      end
+    end
+  end
+end

lib/templates/active_record/migration/create_table_migration.rb

+# See http://doc.gitlab.com/ce/development/migration_style_guide.html
+# for more information on how to write migrations for GitLab.
+
+class <%= migration_class_name %> < ActiveRecord::Migration
+  include Gitlab::Database::MigrationHelpers
+
+  # When using the methods "add_concurrent_index" or "add_column_with_default"
+  # you must disable the use of transactions as these methods can not run in an
+  # existing transaction. When using "add_concurrent_index" make sure that this
+  # method is the _only_ method called in the migration, any other changes
+  # should go in a separate migration. This ensures that upon failure _only_ the
+  # index creation fails and can be retried or reverted easily.
+  #
+  # To disable transactions uncomment the following line and remove these
+  # comments:
+  # disable_ddl_transaction!
+
+  def change
+    create_table :<%= table_name %> do |t|
+<% attributes.each do |attribute| -%>
+<% if attribute.password_digest? -%>
+      t.string :password_digest<%= attribute.inject_options %>
+<% else -%>
+      t.<%= attribute.type %> :<%= attribute.name %><%= attribute.inject_options %>
+<% end -%>
+<% end -%>
+<% if options[:timestamps] %>
+      t.timestamps null: false
+<% end -%>
+    end
+<% attributes_with_index.each do |attribute| -%>
+    add_index :<%= table_name %>, :<%= attribute.index_name %><%= attribute.inject_index_options %>
+<% end -%>
+  end
+end

lib/templates/active_record/migration/migration.rb

+# See http://doc.gitlab.com/ce/development/migration_style_guide.html
+# for more information on how to write migrations for GitLab.
+
+class <%= migration_class_name %> < ActiveRecord::Migration
+  include Gitlab::Database::MigrationHelpers
+
+  # When using the methods "add_concurrent_index" or "add_column_with_default"
+  # you must disable the use of transactions as these methods can not run in an
+  # existing transaction. When using "add_concurrent_index" make sure that this
+  # method is the _only_ method called in the migration, any other changes
+  # should go in a separate migration. This ensures that upon failure _only_ the
+  # index creation fails and can be retried or reverted easily.
+  #
+  # To disable transactions uncomment the following line and remove these
+  # comments:
+  # disable_ddl_transaction!
+
+<%- if migration_action == 'add' -%>
+  def change
+<% attributes.each do |attribute| -%>
+  <%- if attribute.reference? -%>
+    add_reference :<%= table_name %>, :<%= attribute.name %><%= attribute.inject_options %>
+  <%- else -%>
+    add_column :<%= table_name %>, :<%= attribute.name %>, :<%= attribute.type %><%= attribute.inject_options %>
+    <%- if attribute.has_index? -%>
+    add_index :<%= table_name %>, :<%= attribute.index_name %><%= attribute.inject_index_options %>
+    <%- end -%>
+  <%- end -%>
+<%- end -%>
+  end
+<%- elsif migration_action == 'join' -%>
+  def change
+    create_join_table :<%= join_tables.first %>, :<%= join_tables.second %> do |t|
+    <%- attributes.each do |attribute| -%>
+      <%= '# ' unless attribute.has_index? -%>t.index <%= attribute.index_name %><%= attribute.inject_index_options %>
+    <%- end -%>
+    end
+  end
+<%- else -%>
+  def change
+<% attributes.each do |attribute| -%>
+<%- if migration_action -%>
+  <%- if attribute.reference? -%>
+    remove_reference :<%= table_name %>, :<%= attribute.name %><%= attribute.inject_options %>
+  <%- else -%>
+    <%- if attribute.has_index? -%>
+    remove_index :<%= table_name %>, :<%= attribute.index_name %><%= attribute.inject_index_options %>
+    <%- end -%>
+    remove_column :<%= table_name %>, :<%= attribute.name %>, :<%= attribute.type %><%= attribute.inject_options %>
+  <%- end -%>
+<%- end -%>
+<%- end -%>
+  end
+<%- end -%>
+end

spec/lib/gitlab/database/migration_helpers_spec.rb

+require 'spec_helper'
+
+describe Gitlab::Database::MigrationHelpers, lib: true do
+  let(:model) do
+    Class.new do
+      include Gitlab::Database::MigrationHelpers
+
+      def method_missing(name, *args, &block)
+        ActiveRecord::Base.connection.send(name, *args, &block)
+      end
+    end.new
+  end
+
+  describe '#add_concurrent_index' do
+    context 'outside a transaction' do
+      before do
+        expect(model).to receive(:transaction_open?).and_return(false)
+      end
+
+      context 'using PostgreSQL' do
+        it 'creates the index concurrently' do
+          expect(Gitlab::Database).to receive(:postgresql?).and_return(true)
+
+          expect(model).to receive(:add_index).
+            with(:users, :foo, algorithm: :concurrently)
+
+          model.add_concurrent_index(:users, :foo)
+        end
+      end
+
+      context 'using MySQL' do
+        it 'creates a regular index' do
+          expect(Gitlab::Database).to receive(:postgresql?).and_return(false)
+
+          expect(model).to receive(:add_index).
+            with(:users, :foo)
+
+          model.add_concurrent_index(:users, :foo)
+        end
+      end
+    end
+
+    context 'inside a transaction' do
+      it 'raises RuntimeError' do
+        expect(model).to receive(:transaction_open?).and_return(true)
+
+        expect { model.add_concurrent_index(:users, :foo) }.
+          to raise_error(RuntimeError)
+      end
+    end
+  end
+
+  describe '#update_column_in_batches' do
+    before do
+      create_list(:empty_project, 5)
+    end
+
+    it 'updates all the rows in a table' do
+      model.update_column_in_batches(:projects, :import_error, 'foo')
+
+      expect(Project.where(import_error: 'foo').count).to eq(5)
+    end
+  end
+
+  describe '#add_column_with_default' do
+    context 'outside of a transaction' do
+      before do
+        expect(model).to receive(:transaction_open?).and_return(false)
+
+        expect(model).to receive(:transaction).twice.and_yield
+
+        expect(model).to receive(:add_column).
+          with(:projects, :foo, :integer, default: nil)
+
+        expect(model).to receive(:change_column_default).
+          with(:projects, :foo, 10)
+      end
+
+      it 'adds the column while allowing NULL values' do
+        expect(model).to receive(:update_column_in_batches).
+          with(:projects, :foo, 10)
+
+        expect(model).not_to receive(:change_column_null)
+
+        model.add_column_with_default(:projects, :foo, :integer,
+                                      default: 10,
+                                      allow_null: true)
+      end
+
+      it 'adds the column while not allowing NULL values' do
+        expect(model).to receive(:update_column_in_batches).
+          with(:projects, :foo, 10)
+
+        expect(model).to receive(:change_column_null).
+          with(:projects, :foo, false)
+
+        model.add_column_with_default(:projects, :foo, :integer, default: 10)
+      end
+
+      it 'removes the added column whenever updating the rows fails' do
+        expect(model).to receive(:update_column_in_batches).
+          with(:projects, :foo, 10).
+          and_raise(RuntimeError)
+
+        expect(model).to receive(:remove_column).
+          with(:projects, :foo)
+
+        expect do
+          model.add_column_with_default(:projects, :foo, :integer, default: 10)
+        end.to raise_error(RuntimeError)
+      end
+    end
+
+    context 'inside a transaction' do
+      it 'raises RuntimeError' do
+        expect(model).to receive(:transaction_open?).and_return(true)
+
+        expect do
+          model.add_column_with_default(:projects, :foo, :integer, default: 10)
+        end.to raise_error(RuntimeError)
+      end
+    end
+  end
+end