flaky

About

Flaky is a plugin for nose or py.test that automatically reruns flaky tests.

Ideally, tests reliably pass or fail, but sometimes test fixtures must rely on components that aren't 100% reliable. With flaky, instead of removing those tests or marking them to \@skip, they can be automatically retried.

For more information about flaky, see this presentation.

Marking tests flaky

To mark a test as flaky, simply import flaky and decorate the test with \@flaky:

from flaky import flaky

@flaky
def test_something_that_usually_passes(self):
    value_to_double = 21
    result = get_result_from_flaky_doubler(value_to_double)
    self.assertEqual(result, value_to_double * 2, 'Result doubled incorrectly.')

By default, flaky will retry a failing test once, but that behavior can be overridden by passing values to the flaky decorator. It accepts two parameters: max_runs, and min_passes; flaky will run tests up to max_runs times, until it has succeeded min_passes times. Once a test passes min_passes times, it's considered a success; once it has been run max_runs times without passing min_passes times, it's considered a failure.

@flaky(max_runs=3, min_passes=2)
def test_something_that_usually_passes(self):
    """This test must pass twice, and it can be run up to three times."""
    value_to_double = 21
    result = get_result_from_flaky_doubler(value_to_double)
    self.assertEqual(result, value_to_double * 2, 'Result doubled incorrectly.')

Marking a class flaky

In addition to marking a single test flaky, entire test cases can be marked flaky:

@flaky
class TestMultipliers(TestCase):
    def test_flaky_doubler(self):
        value_to_double = 21
        result = get_result_from_flaky_doubler(value_to_double)
        self.assertEqual(result, value_to_double * 2, 'Result doubled incorrectly.')

    @flaky(max_runs=3)
    def test_flaky_tripler(self):
        value_to_triple = 14
        result = get_result_from_flaky_tripler(value_to_triple)
        self.assertEqual(result, value_to_triple * 3, 'Result tripled incorrectly.')

The \@flaky class decorator will mark test_flaky_doubler as flaky, but it won't override the 3 max_runs for test_flaky_tripler (from the decorator on that test method).

Don't rerun certain types of failures

Depending on your tests, some failures are obviously not due to flakiness. Instead of rerunning after those failures, you can specify a filter function that can tell flaky to fail the test right away.

def is_not_crash(err, *args):
    return not issubclass(err[0], ProductCrashedError)

@flaky
def test_something():
    raise ProductCrashedError

@flaky(rerun_filter=is_not_crash)
def test_something_else():
    raise ProductCrashedError

File truncated at 100 lines see the full file