Intermittent ActiveRecord::RecordNotFound during Ruby on Rails Integration Tests

Question

I have integration tests that are intermittently failing, and always with ActiveRecord::RecordNotFound errors. The error takes place inside the controller where a find call takes place given an ID from a fixture. It takes place in multiple controllers, though. I never see this behaviour while navigating the site, but I'd say the tests fail about 30-50% of the time. Running the test again after a failure seems to fix the problem.

If I load the fixtures manually into the development database, the ID that doesn't seem to be found is indeed present inside the tables.

I haven't been able to find much info on people having the same problem... any ideas?

UPDATE: Here are the contents of test_helper.rb

ENV["RAILS_ENV"] = "test"
require File.expand_path('../../config/environment', __FILE__)
require 'rails/test_help'
require 'capybara/rails'
class ActiveSupport::TestCase
  # Setup all fixtures in test/fixtures/*.(yml|csv) for all tests in alphabetical order.
  #
  # Note: You'll currently still have to declare fixtures explicitly in integration tests
  # -- they do not yet inherit this setting
  fixtures :all
  # Add more helper methods to be used by all tests here...
end
# Transactional fixtures do not work with Selenium tests, because Capybara
# uses a separate server thread, which the transactions would be hidden
# from. We hence use DatabaseCleaner to truncate our test database.
DatabaseCleaner.strategy = :truncation
class ActionDispatch::IntegrationTest
  # Make the Capybara DSL available in all integration tests
  include Capybara::DSL
  # Make the Capybara Email DSL available in all integration tests
  include Capybara::Email::DSL
  # Stop ActiveRecord from wrapping tests in transactions
  self.use_transactional_fixtures = false
  # Switch to selenium as the default driver for JS support
  Capybara.default_driver = :selenium
  # Only click on visible links!
  Capybara.ignore_hidden_elements = true
  teardown do
    DatabaseCleaner.clean       # Truncate the database
    Capybara.reset_sessions!    # Forget the (simulated) browser state
    Capybara.use_default_driver # Revert Capybara.current_driver to Capybara.default_driver
  end
end

UPDATE Here are the results of running the same test 5 times in a row. I ran the tests, waited until they finished, and then immediately ran them again with the command rails test:integration. NOTE: The E and F that are consistent throughout the tests are actually test errors -- I'm working on fixing those. For example, the second test run-through was "correct", but the first showed the spurious error.

..E......E..........F.
.........E..........F.
..E......E..........F.
..E......E..........F.
..E....E.E..........F.

The errors do occur across two separate tables -- they're not trying to find the same record. But it does appear to be only a subset of the tests that have this issue...

UPDATE Here's what the actual error looks like in the test results:

  1) Error:
test_browsing_user_snops(BrowseStoriesTest):
ActiveRecord::RecordNotFound: Couldn't find User with id=980190962
    /home/myuser/.rvm/gems/ruby-1.9.3-p286/gems/activerecord-3.2.9/lib/active_record/relation/finder_methods.rb:341:in `find_one'
    /home/myuser/.rvm/gems/ruby-1.9.3-p286/gems/activerecord-3.2.9/lib/active_record/relation/finder_methods.rb:312:in `find_with_ids'
    /home/myuser/.rvm/gems/ruby-1.9.3-p286/gems/activerecord-3.2.9/lib/active_record/relation/finder_methods.rb:107:in `find'
    /home/myuser/.rvm/gems/ruby-1.9.3-p286/gems/activerecord-3.2.9/lib/active_record/querying.rb:5:in `find'
    /home/myuser/Projects/myproject/app/controllers/users_controller.rb:18:in `show'
    ...

Notice that the error appears when the controller tries to find a record. The relevant line is actually @user = User.find(params[:id]). It occurs with other models also, not just the users controller and not just the User model.

Answer 1

I'm concerned that there may be a delay during the truncation of data and the speed in which Capybara is driving the web browser which sometimes may result in identity column values starting at an unexpected value (i.e., 7 for record #1 - when you expect 1, because that identity generator has not yet been reset). I have no evidence that this is the case but its my best guess.

Take at look at item #3 at this URL. This is a pretty straight-forward hack, set use_transactional_fixtures to true and monkeypatch ActiveRecord by pasting that code into your test_helper.rb. This could help in eliminating any intermittent disk IO problems that could be a potential problem.

Another thing you can try is restricting your SQLite test database by setting the filename for that database to :memory: in your database.yml file. This should accomplish the same thing as above - eliminate spurious disk IO that may be causing these intermittent issues.