Journey of a complex gem upgrade

Transcript

1. Journey of a complex gem upgrade
2. I ❤ Japan!
3. Edouard Chin https://github.com/Edouard-chin @DaroudeDudek
4. Make the upgrade process faster, smoother and safer
5. Works for any gem upgrade
6. 1. Spot breaking changes in the CHANGELOG 2. `bundle update` 3. Deploy
8. 1. Spot breaking changes in the CHANGELOG 2. `bundle update` 3. Deploy
10. Progressively fix all issues
11. Easy to upgrade and solely used in test/development Tough to upgrade, takes a long time and effort Relatively easy to bump but critical for your infra
12. Running in production since 2006 600.000 Merchants One of the largest Ruby application
13. $ bundle-stats
15. Upgrade can be delayed by months without proper tooling
16. Full upgrade cycle ⏱
 
 6 months - 1 year
17. 1. Prepare our application for dual boot
18. source 'https://rubygems.org' gem 'rubykaigi', ‘3.0’
19. source 'https://rubygems.org' if ENV[‘SHOPIFY_NEXT’] bundler_monkey_patch end if ENV['SHOPIFY_NEXT'] gem 'rubykaigi', ‘3.1’ else gem 'rubykaigi', ‘3.0’ end
21. 2. Fix all issues that might happen on boot
22. 2868 errors
25. 2868 errors
27. Lessons learnt: Stop the bleeding first
28. 1. Prepare app for dual-booting 2. Fix issues happening on boot 3. Enable CI
29. Make CI fail only when new broken code is introduced Existing broken tests will be marked and allowed to fail
30. mark_as :failing_on_rails_next test 'example test' do raise StandardError, ‘This test is marked as failing’ end
31. mark_as :failing_on_rails_next test 'example test' do raise StandardError, 'This test fails but will be considered as passing' end # What's the difference between both test 'example test' do skip('This test is failing lets not run it') if running_on_rails_next? raise StandardError, 'This test fails but will be skipped' end
32. mark_as :failing_on_rails_next test 'example test' do raise StandardError, ‘This test is marked as failing’ end desc 'This task is great' task :great_task do # ... end desc 'Another cool task' task :cool_task do # ... end
33. # frozen_string_literal: true class MyTest < MiniTest::Test include MarkingModule mark_as :failing_on_rails_next def test_example … end end puts MyTest.new('test_example').failing_on_rails_next? # true
34. Minitest Reporter
35. ProgressReporter SummaryReporter
36. # Snippet taken from minitest.rb at_exit { exit exit_code || false } exit_code = Minitest.run ARGV
37. module Minitest module Reporters class RailsNextReporter < BaseReporter ALLOWED = :allowed attr_writer :state def after_test(test) return unless test_will_fail?(test) @state = ALLOWED end def test_will_fail?(test) return false if not_on_rails_next test.failing_on_rails_next? end def record(result) result.failures.clear if @state == ALLOWED end end end end
40. Side advantage on stopping the bleeding: Getting help from other developers.
41. Workforce )*+,-). +,.)*. *+,+,.,. )*+,-). +,.)*. *+,+,.,.
42. Componentization
43. Componentization Before After
44. Natural way to identify which code belongs to which component.
46. ⛵
50. The less deprecations, the easier it will be to upgrade
52. ActiveSupport::Deprecation.warn(‘foo is deprecated’)
53. You have introduced new deprecations in the codebase
55. — test_foo: - DEPRECATION WARNING: This code is deprecated - DEPRECATION WARNING: Another deprecation test_bar - DEPRECATION WARNING: This code is deprecated class MyTest < MiniTest::Test def test_foo ActiveSupport::Deprecation.warn(‘This code is deprecated') ActiveSupport::Deprecation.warn(‘Another deprecation’) end def test_bar ActiveSupport::Deprecation.warn(‘This code is deprecated’) end end
57. 1. Add rotating seats in planes
58. 2. Have a way to run your app with different version of a gem
59. 3. Stop the bleeding
60. 4. Have a clear distinction on who should own what and parallelize the work 5. Add some gamification
61. 6. Deployment tooling to mitigate the risk
62. 7. Keep track and fix deprecations ASAP
63. Edouard Chin https://github.com/Edouard-chin @DaroudeDudek Thanks!