class AssistantTest(unittest.TestCase):
def setUp(self):
self.sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
self.w = Worker(scheduler=self.sch, worker_id='X').__enter__()
self.assistant = Worker(scheduler=self.sch,
worker_id='Y',
assistant=True)
def tearDown(self):
self.w.__exit__(None, None, None)
def test_get_work(self):
d = Dummy2Task('123')
self.w.add(d)
self.assertFalse(d.complete())
self.assistant.run()
self.assertTrue(d.complete())
def test_bad_job_type(self):
class Dummy3Task(Dummy2Task):
task_family = 'UnknownTaskFamily'
d = Dummy3Task('123')
self.w.add(d)
self.assertFalse(d.complete())
self.assertFalse(self.assistant.run())
self.assertFalse(d.complete())
self.assertEqual(list(self.sch.task_list('FAILED', '').keys()),
[str(d)])
def test_unimported_job_type(self):
class NotImportedTask(luigi.Task):
task_family = 'UnimportedTask'
task_module = None
task = NotImportedTask()
# verify that it can't run the task without the module info necessary to import it
self.w.add(task)
self.assertFalse(self.assistant.run())
self.assertEqual(list(self.sch.task_list('FAILED', '').keys()),
['UnimportedTask()'])
# check that it can import with the right module
task.task_module = 'dummy_test_module.not_imported'
self.w.add(task)
self.assertTrue(self.assistant.run())
self.assertEqual(list(self.sch.task_list('DONE', '').keys()),
['UnimportedTask()'])
class AssistantTest(unittest.TestCase):
def setUp(self):
self.sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
self.w = Worker(scheduler=self.sch, worker_id='X').__enter__()
self.assistant = Worker(scheduler=self.sch, worker_id='Y', assistant=True)
def tearDown(self):
self.w.__exit__(None, None, None)
def test_get_work(self):
d = Dummy2Task('123')
self.w.add(d)
self.assertFalse(d.complete())
self.assistant.run()
self.assertTrue(d.complete())
def test_bad_job_type(self):
class Dummy3Task(Dummy2Task):
task_family = 'UnknownTaskFamily'
d = Dummy3Task('123')
self.w.add(d)
self.assertFalse(d.complete())
self.assertFalse(self.assistant.run())
self.assertFalse(d.complete())
self.assertEqual(list(self.sch.task_list('FAILED', '').keys()), [str(d)])
def test_unimported_job_type(self):
class NotImportedTask(luigi.Task):
task_family = 'UnimportedTask'
task_module = None
task = NotImportedTask()
# verify that it can't run the task without the module info necessary to import it
self.w.add(task)
self.assertFalse(self.assistant.run())
self.assertEqual(list(self.sch.task_list('FAILED', '').keys()), ['UnimportedTask()'])
# check that it can import with the right module
task.task_module = 'dummy_test_module.not_imported'
self.w.add(task)
self.assertTrue(self.assistant.run())
self.assertEqual(list(self.sch.task_list('DONE', '').keys()), ['UnimportedTask()'])
class WorkerEmailTest(unittest.TestCase):
def setUp(self):
super(WorkerEmailTest, self).setUp()
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
self.worker = Worker(scheduler=sch, worker_id="foo").__enter__()
def tearDown(self):
self.worker.__exit__(None, None, None)
@email_patch
def test_connection_error(self, emails):
sch = RemoteScheduler('http://tld.invalid:1337', connect_timeout=1)
self.waits = 0
def dummy_wait():
self.waits += 1
sch._wait = dummy_wait
class A(DummyTask):
pass
a = A()
self.assertEqual(emails, [])
with Worker(scheduler=sch) as worker:
worker.add(a)
self.assertEqual(self.waits, 2) # should attempt to add it 3 times
self.assertNotEqual(emails, [])
self.assertTrue(
emails[0].find("Luigi: Framework error while scheduling %s" %
(a, )) != -1)
@email_patch
def test_complete_error(self, emails):
class A(DummyTask):
def complete(self):
raise Exception("b0rk")
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" %
(a, )) != -1)
self.worker.run()
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" %
(a, )) != -1)
self.assertFalse(a.has_run)
@email_patch
def test_requires_error(self, emails):
class A(DummyTask):
def requires(self):
raise Exception("b0rk")
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" %
(a, )) != -1)
self.worker.run()
self.assertFalse(a.has_run)
@email_patch
def test_complete_return_value(self, emails):
class A(DummyTask):
def complete(self):
pass # no return value should be an error
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" %
(a, )) != -1)
self.worker.run()
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" %
(a, )) != -1)
self.assertFalse(a.has_run)
@email_patch
def test_run_error(self, emails):
class A(luigi.Task):
def complete(self):
return False
def run(self):
raise Exception("b0rk")
a = A()
self.worker.add(a)
self.assertEqual(emails, [])
self.worker.run()
self.assertTrue(emails[0].find("Luigi: %s FAILED" % (a, )) != -1)
@email_patch
def test_no_error(self, emails):
class A(DummyTask):
pass
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertEqual(emails, [])
self.worker.run()
self.assertEqual(emails, [])
self.assertTrue(a.complete())
@custom_email_patch({
"core": {
"error-email": "not-a-real-email-address-for-test-only",
'email-type': 'none'
}
})
def test_disable_emails(self, emails):
class A(luigi.Task):
def complete(self):
raise Exception("b0rk")
self.worker.add(A())
self.assertEqual(emails, [])
class WorkerTest(unittest.TestCase):
def setUp(self):
# InstanceCache.disable()
self.sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
self.w = Worker(scheduler=self.sch, worker_id='X').__enter__()
self.w2 = Worker(scheduler=self.sch, worker_id='Y').__enter__()
self.time = time.time
def tearDown(self):
if time.time != self.time:
time.time = self.time
self.w.__exit__(None, None, None)
self.w2.__exit__(None, None, None)
def setTime(self, t):
time.time = lambda: t
def test_dep(self):
class A(Task):
def run(self):
self.has_run = True
def complete(self):
return self.has_run
a = A()
class B(Task):
def requires(self):
return a
def run(self):
self.has_run = True
def complete(self):
return self.has_run
b = B()
a.has_run = False
b.has_run = False
self.assertTrue(self.w.add(b))
self.assertTrue(self.w.run())
self.assertTrue(a.has_run)
self.assertTrue(b.has_run)
def test_stop_getting_new_work(self):
d = DummyTask()
self.w.add(d)
self.assertFalse(d.complete())
try:
self.w.handle_interrupt(signal.SIGUSR1, None)
except AttributeError:
raise unittest.SkipTest('signal.SIGUSR1 not found on this system')
self.w.run()
self.assertFalse(d.complete())
def test_external_dep(self):
class A(ExternalTask):
def complete(self):
return False
a = A()
class B(Task):
def requires(self):
return a
def run(self):
self.has_run = True
def complete(self):
return self.has_run
b = B()
a.has_run = False
b.has_run = False
self.assertTrue(self.w.add(b))
self.assertTrue(self.w.run())
self.assertFalse(a.has_run)
self.assertFalse(b.has_run)
def test_tracking_url(self):
tracking_url = 'http://test_url.com/'
class A(Task):
has_run = False
def complete(self):
return self.has_run
def run(self, tracking_url_callback=None):
if tracking_url_callback is not None:
tracking_url_callback(tracking_url)
self.has_run = True
a = A()
self.assertTrue(self.w.add(a))
self.assertTrue(self.w.run())
tasks = self.sch.task_list('DONE', '')
self.assertEqual(1, len(tasks))
self.assertEqual(tracking_url, tasks['A()']['tracking_url'])
def test_type_error_in_tracking_run(self):
class A(Task):
num_runs = 0
def complete(self):
return False
def run(self, tracking_url_callback=None):
self.num_runs += 1
raise TypeError('bad type')
a = A()
self.assertTrue(self.w.add(a))
self.assertFalse(self.w.run())
# Should only run and fail once, not retry because of the type error
self.assertEqual(1, a.num_runs)
def test_fail(self):
class CustomException(BaseException):
def __init__(self, msg):
self.msg = msg
class A(Task):
def run(self):
self.has_run = True
raise CustomException('bad things')
def complete(self):
return self.has_run
a = A()
class B(Task):
def requires(self):
return a
def run(self):
self.has_run = True
def complete(self):
return self.has_run
b = B()
a.has_run = False
b.has_run = False
self.assertTrue(self.w.add(b))
self.assertFalse(self.w.run())
self.assertTrue(a.has_run)
self.assertFalse(b.has_run)
def test_unknown_dep(self):
# see central_planner_test.CentralPlannerTest.test_remove_dep
class A(ExternalTask):
def complete(self):
return False
class C(Task):
def complete(self):
return True
def get_b(dep):
class B(Task):
def requires(self):
return dep
def run(self):
self.has_run = True
def complete(self):
return False
b = B()
b.has_run = False
return b
b_a = get_b(A())
b_c = get_b(C())
self.assertTrue(self.w.add(b_a))
# So now another worker goes in and schedules C -> B
# This should remove the dep A -> B but will screw up the first worker
self.assertTrue(self.w2.add(b_c))
self.assertFalse(
self.w.run()
) # should not run anything - the worker should detect that A is broken
self.assertFalse(b_a.has_run)
# not sure what should happen??
# self.w2.run() # should run B since C is fulfilled
# self.assertTrue(b_c.has_run)
def test_unfulfilled_dep(self):
class A(Task):
def complete(self):
return self.done
def run(self):
self.done = True
def get_b(a):
class B(A):
def requires(self):
return a
b = B()
b.done = False
a.done = True
return b
a = A()
b = get_b(a)
self.assertTrue(self.w.add(b))
a.done = False
self.w.run()
self.assertTrue(a.complete())
self.assertTrue(b.complete())
def test_gets_missed_work(self):
class A(Task):
done = False
def complete(self):
return self.done
def run(self):
self.done = True
a = A()
self.assertTrue(self.w.add(a))
# simulate a missed get_work response
self.assertEqual('A()', self.sch.get_work(worker='X')['task_id'])
self.assertTrue(self.w.run())
self.assertTrue(a.complete())
def test_avoid_infinite_reschedule(self):
class A(Task):
def complete(self):
return False
class B(Task):
def complete(self):
return False
def requires(self):
return A()
self.assertTrue(self.w.add(B()))
self.assertFalse(self.w.run())
def test_fails_registering_signal(self):
with mock.patch('luigi.worker.signal', spec=['signal']):
# mock will raise an attribute error getting signal.SIGUSR1
Worker()
def test_allow_reschedule_with_many_missing_deps(self):
class A(Task):
""" Task that must run twice to succeed """
i = luigi.IntParameter()
runs = 0
def complete(self):
return self.runs >= 2
def run(self):
self.runs += 1
class B(Task):
done = False
def requires(self):
return map(A, range(20))
def complete(self):
return self.done
def run(self):
self.done = True
b = B()
w = Worker(scheduler=self.sch, worker_id='X', max_reschedules=1)
self.assertTrue(w.add(b))
self.assertFalse(w.run())
# For b to be done, we must have rescheduled its dependencies to run them twice
self.assertTrue(b.complete())
self.assertTrue(all(a.complete() for a in b.deps()))
def test_interleaved_workers(self):
class A(DummyTask):
pass
a = A()
class B(DummyTask):
def requires(self):
return a
class ExternalB(ExternalTask):
task_family = "B"
def complete(self):
return False
b = B()
eb = ExternalB()
self.assertEqual(eb.task_id, "B()")
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
with Worker(scheduler=sch,
worker_id='X') as w, Worker(scheduler=sch,
worker_id='Y') as w2:
self.assertTrue(w.add(b))
self.assertTrue(w2.add(eb))
logging.debug("RUNNING BROKEN WORKER")
self.assertTrue(w2.run())
self.assertFalse(a.complete())
self.assertFalse(b.complete())
logging.debug("RUNNING FUNCTIONAL WORKER")
self.assertTrue(w.run())
self.assertTrue(a.complete())
self.assertTrue(b.complete())
def test_interleaved_workers2(self):
# two tasks without dependencies, one external, one not
class B(DummyTask):
pass
class ExternalB(ExternalTask):
task_family = "B"
def complete(self):
return False
b = B()
eb = ExternalB()
self.assertEqual(eb.task_id, "B()")
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
with Worker(scheduler=sch,
worker_id='X') as w, Worker(scheduler=sch,
worker_id='Y') as w2:
self.assertTrue(w2.add(eb))
self.assertTrue(w.add(b))
self.assertTrue(w2.run())
self.assertFalse(b.complete())
self.assertTrue(w.run())
self.assertTrue(b.complete())
def test_interleaved_workers3(self):
class A(DummyTask):
def run(self):
logging.debug('running A')
time.sleep(0.1)
super(A, self).run()
a = A()
class B(DummyTask):
def requires(self):
return a
def run(self):
logging.debug('running B')
super(B, self).run()
b = B()
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
with Worker(scheduler=sch,
worker_id='X',
keep_alive=True,
count_uniques=True) as w:
with Worker(scheduler=sch,
worker_id='Y',
keep_alive=True,
count_uniques=True,
wait_interval=0.1) as w2:
self.assertTrue(w.add(a))
self.assertTrue(w2.add(b))
threading.Thread(target=w.run).start()
self.assertTrue(w2.run())
self.assertTrue(a.complete())
self.assertTrue(b.complete())
def test_die_for_non_unique_pending(self):
class A(DummyTask):
def run(self):
logging.debug('running A')
time.sleep(0.1)
super(A, self).run()
a = A()
class B(DummyTask):
def requires(self):
return a
def run(self):
logging.debug('running B')
super(B, self).run()
b = B()
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
with Worker(scheduler=sch,
worker_id='X',
keep_alive=True,
count_uniques=True) as w:
with Worker(scheduler=sch,
worker_id='Y',
keep_alive=True,
count_uniques=True,
wait_interval=0.1) as w2:
self.assertTrue(w.add(b))
self.assertTrue(w2.add(b))
self.assertEqual(w._get_work()[0], 'A()')
self.assertTrue(w2.run())
self.assertFalse(a.complete())
self.assertFalse(b.complete())
def test_complete_exception(self):
"Tests that a task is still scheduled if its sister task crashes in the complete() method"
class A(DummyTask):
def complete(self):
raise Exception("doh")
a = A()
class C(DummyTask):
pass
c = C()
class B(DummyTask):
def requires(self):
return a, c
b = B()
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id="foo") as w:
self.assertFalse(w.add(b))
self.assertTrue(w.run())
self.assertFalse(b.has_run)
self.assertTrue(c.has_run)
self.assertFalse(a.has_run)
def test_requires_exception(self):
class A(DummyTask):
def requires(self):
raise Exception("doh")
a = A()
class D(DummyTask):
pass
d = D()
class C(DummyTask):
def requires(self):
return d
c = C()
class B(DummyTask):
def requires(self):
return c, a
b = B()
sch = CentralPlannerScheduler(retry_delay=100,
remove_delay=1000,
worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id="foo") as w:
self.assertFalse(w.add(b))
self.assertTrue(w.run())
self.assertFalse(b.has_run)
self.assertTrue(c.has_run)
self.assertTrue(d.has_run)
self.assertFalse(a.has_run)
class MultiprocessWorkerTest(unittest.TestCase):
def setUp(self):
self.scheduler = RemoteScheduler()
self.scheduler.add_worker = Mock()
self.scheduler.add_task = Mock()
self.worker = Worker(scheduler=self.scheduler, worker_id='X', worker_processes=2).__enter__()
def tearDown(self):
self.worker.__exit__(None, None, None)
def gw_res(self, pending, task_id):
return dict(n_pending_tasks=pending,
task_id=task_id,
running_tasks=0, n_unique_pending=0)
def test_positive_path(self):
a = DummyTask("a")
b = DummyTask("b")
class MultipleRequirementTask(DummyTask):
def requires(self):
return [a, b]
c = MultipleRequirementTask("C")
self.assertTrue(self.worker.add(c))
self.scheduler.get_work = Mock(side_effect=[self.gw_res(3, str(a)),
self.gw_res(2, str(b)),
self.gw_res(1, str(c)),
self.gw_res(0, None),
self.gw_res(0, None)])
self.assertTrue(self.worker.run())
self.assertTrue(c.has_run)
def test_path_with_task_failures(self):
class FailingTask(DummyTask):
def run(self):
raise Exception("I am failing")
a = FailingTask("a")
b = FailingTask("b")
class MultipleRequirementTask(DummyTask):
def requires(self):
return [a, b]
c = MultipleRequirementTask("C")
self.assertTrue(self.worker.add(c))
self.scheduler.get_work = Mock(side_effect=[self.gw_res(3, str(a)),
self.gw_res(2, str(b)),
self.gw_res(1, str(c)),
self.gw_res(0, None),
self.gw_res(0, None)])
self.assertFalse(self.worker.run())
class WorkerEmailTest(unittest.TestCase):
def setUp(self):
super(WorkerEmailTest, self).setUp()
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
self.worker = Worker(scheduler=sch, worker_id="foo").__enter__()
def tearDown(self):
self.worker.__exit__(None, None, None)
@email_patch
def test_connection_error(self, emails):
sch = RemoteScheduler('http://tld.invalid:1337', connect_timeout=1)
self.waits = 0
def dummy_wait():
self.waits += 1
sch._wait = dummy_wait
class A(DummyTask):
pass
a = A()
self.assertEqual(emails, [])
with Worker(scheduler=sch) as worker:
worker.add(a)
self.assertEqual(self.waits, 2) # should attempt to add it 3 times
self.assertNotEqual(emails, [])
self.assertTrue(emails[0].find("Luigi: Framework error while scheduling %s" % (a,)) != -1)
@email_patch
def test_complete_error(self, emails):
class A(DummyTask):
def complete(self):
raise Exception("b0rk")
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" % (a,)) != -1)
self.worker.run()
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" % (a,)) != -1)
self.assertFalse(a.has_run)
@email_patch
def test_requires_error(self, emails):
class A(DummyTask):
def requires(self):
raise Exception("b0rk")
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" % (a,)) != -1)
self.worker.run()
self.assertFalse(a.has_run)
@email_patch
def test_complete_return_value(self, emails):
class A(DummyTask):
def complete(self):
pass # no return value should be an error
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" % (a,)) != -1)
self.worker.run()
self.assertTrue(emails[0].find("Luigi: %s failed scheduling" % (a,)) != -1)
self.assertFalse(a.has_run)
@email_patch
def test_run_error(self, emails):
class A(luigi.Task):
def complete(self):
return False
def run(self):
raise Exception("b0rk")
a = A()
self.worker.add(a)
self.assertEqual(emails, [])
self.worker.run()
self.assertTrue(emails[0].find("Luigi: %s FAILED" % (a,)) != -1)
@email_patch
def test_no_error(self, emails):
class A(DummyTask):
pass
a = A()
self.assertEqual(emails, [])
self.worker.add(a)
self.assertEqual(emails, [])
self.worker.run()
self.assertEqual(emails, [])
self.assertTrue(a.complete())
@custom_email_patch({"core": {"error-email": "not-a-real-email-address-for-test-only", 'email-type': 'none'}})
def test_disable_emails(self, emails):
class A(luigi.Task):
def complete(self):
raise Exception("b0rk")
self.worker.add(A())
self.assertEqual(emails, [])
class WorkerTest(unittest.TestCase):
def setUp(self):
# InstanceCache.disable()
self.sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
self.w = Worker(scheduler=self.sch, worker_id='X').__enter__()
self.w2 = Worker(scheduler=self.sch, worker_id='Y').__enter__()
self.time = time.time
def tearDown(self):
if time.time != self.time:
time.time = self.time
self.w.__exit__(None, None, None)
self.w2.__exit__(None, None, None)
def setTime(self, t):
time.time = lambda: t
def test_dep(self):
class A(Task):
def run(self):
self.has_run = True
def complete(self):
return self.has_run
a = A()
class B(Task):
def requires(self):
return a
def run(self):
self.has_run = True
def complete(self):
return self.has_run
b = B()
a.has_run = False
b.has_run = False
self.assertTrue(self.w.add(b))
self.assertTrue(self.w.run())
self.assertTrue(a.has_run)
self.assertTrue(b.has_run)
def test_stop_getting_new_work(self):
d = DummyTask()
self.w.add(d)
self.assertFalse(d.complete())
try:
self.w.handle_interrupt(signal.SIGUSR1, None)
except AttributeError:
raise unittest.SkipTest('signal.SIGUSR1 not found on this system')
self.w.run()
self.assertFalse(d.complete())
def test_external_dep(self):
class A(ExternalTask):
def complete(self):
return False
a = A()
class B(Task):
def requires(self):
return a
def run(self):
self.has_run = True
def complete(self):
return self.has_run
b = B()
a.has_run = False
b.has_run = False
self.assertTrue(self.w.add(b))
self.assertTrue(self.w.run())
self.assertFalse(a.has_run)
self.assertFalse(b.has_run)
def test_tracking_url(self):
tracking_url = 'http://test_url.com/'
class A(Task):
has_run = False
def complete(self):
return self.has_run
def run(self, tracking_url_callback=None):
if tracking_url_callback is not None:
tracking_url_callback(tracking_url)
self.has_run = True
a = A()
self.assertTrue(self.w.add(a))
self.assertTrue(self.w.run())
tasks = self.sch.task_list('DONE', '')
self.assertEqual(1, len(tasks))
self.assertEqual(tracking_url, tasks['A()']['tracking_url'])
def test_type_error_in_tracking_run(self):
class A(Task):
num_runs = 0
def complete(self):
return False
def run(self, tracking_url_callback=None):
self.num_runs += 1
raise TypeError('bad type')
a = A()
self.assertTrue(self.w.add(a))
self.assertFalse(self.w.run())
# Should only run and fail once, not retry because of the type error
self.assertEqual(1, a.num_runs)
def test_fail(self):
class CustomException(BaseException):
def __init__(self, msg):
self.msg = msg
class A(Task):
def run(self):
self.has_run = True
raise CustomException('bad things')
def complete(self):
return self.has_run
a = A()
class B(Task):
def requires(self):
return a
def run(self):
self.has_run = True
def complete(self):
return self.has_run
b = B()
a.has_run = False
b.has_run = False
self.assertTrue(self.w.add(b))
self.assertFalse(self.w.run())
self.assertTrue(a.has_run)
self.assertFalse(b.has_run)
def test_unknown_dep(self):
# see central_planner_test.CentralPlannerTest.test_remove_dep
class A(ExternalTask):
def complete(self):
return False
class C(Task):
def complete(self):
return True
def get_b(dep):
class B(Task):
def requires(self):
return dep
def run(self):
self.has_run = True
def complete(self):
return False
b = B()
b.has_run = False
return b
b_a = get_b(A())
b_c = get_b(C())
self.assertTrue(self.w.add(b_a))
# So now another worker goes in and schedules C -> B
# This should remove the dep A -> B but will screw up the first worker
self.assertTrue(self.w2.add(b_c))
self.assertFalse(self.w.run()) # should not run anything - the worker should detect that A is broken
self.assertFalse(b_a.has_run)
# not sure what should happen??
# self.w2.run() # should run B since C is fulfilled
# self.assertTrue(b_c.has_run)
def test_unfulfilled_dep(self):
class A(Task):
def complete(self):
return self.done
def run(self):
self.done = True
def get_b(a):
class B(A):
def requires(self):
return a
b = B()
b.done = False
a.done = True
return b
a = A()
b = get_b(a)
self.assertTrue(self.w.add(b))
a.done = False
self.w.run()
self.assertTrue(a.complete())
self.assertTrue(b.complete())
def test_gets_missed_work(self):
class A(Task):
done = False
def complete(self):
return self.done
def run(self):
self.done = True
a = A()
self.assertTrue(self.w.add(a))
# simulate a missed get_work response
self.assertEqual('A()', self.sch.get_work(worker='X')['task_id'])
self.assertTrue(self.w.run())
self.assertTrue(a.complete())
def test_avoid_infinite_reschedule(self):
class A(Task):
def complete(self):
return False
class B(Task):
def complete(self):
return False
def requires(self):
return A()
self.assertTrue(self.w.add(B()))
self.assertFalse(self.w.run())
def test_fails_registering_signal(self):
with mock.patch('luigi.worker.signal', spec=['signal']):
# mock will raise an attribute error getting signal.SIGUSR1
Worker()
def test_allow_reschedule_with_many_missing_deps(self):
class A(Task):
""" Task that must run twice to succeed """
i = luigi.IntParameter()
runs = 0
def complete(self):
return self.runs >= 2
def run(self):
self.runs += 1
class B(Task):
done = False
def requires(self):
return map(A, range(20))
def complete(self):
return self.done
def run(self):
self.done = True
b = B()
w = Worker(scheduler=self.sch, worker_id='X', max_reschedules=1)
self.assertTrue(w.add(b))
self.assertFalse(w.run())
# For b to be done, we must have rescheduled its dependencies to run them twice
self.assertTrue(b.complete())
self.assertTrue(all(a.complete() for a in b.deps()))
def test_interleaved_workers(self):
class A(DummyTask):
pass
a = A()
class B(DummyTask):
def requires(self):
return a
class ExternalB(ExternalTask):
task_family = "B"
def complete(self):
return False
b = B()
eb = ExternalB()
self.assertEqual(eb.task_id, "B()")
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id='X') as w, Worker(scheduler=sch, worker_id='Y') as w2:
self.assertTrue(w.add(b))
self.assertTrue(w2.add(eb))
logging.debug("RUNNING BROKEN WORKER")
self.assertTrue(w2.run())
self.assertFalse(a.complete())
self.assertFalse(b.complete())
logging.debug("RUNNING FUNCTIONAL WORKER")
self.assertTrue(w.run())
self.assertTrue(a.complete())
self.assertTrue(b.complete())
def test_interleaved_workers2(self):
# two tasks without dependencies, one external, one not
class B(DummyTask):
pass
class ExternalB(ExternalTask):
task_family = "B"
def complete(self):
return False
b = B()
eb = ExternalB()
self.assertEqual(eb.task_id, "B()")
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id='X') as w, Worker(scheduler=sch, worker_id='Y') as w2:
self.assertTrue(w2.add(eb))
self.assertTrue(w.add(b))
self.assertTrue(w2.run())
self.assertFalse(b.complete())
self.assertTrue(w.run())
self.assertTrue(b.complete())
def test_interleaved_workers3(self):
class A(DummyTask):
def run(self):
logging.debug('running A')
time.sleep(0.1)
super(A, self).run()
a = A()
class B(DummyTask):
def requires(self):
return a
def run(self):
logging.debug('running B')
super(B, self).run()
b = B()
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id='X', keep_alive=True, count_uniques=True) as w:
with Worker(scheduler=sch, worker_id='Y', keep_alive=True, count_uniques=True, wait_interval=0.1) as w2:
self.assertTrue(w.add(a))
self.assertTrue(w2.add(b))
threading.Thread(target=w.run).start()
self.assertTrue(w2.run())
self.assertTrue(a.complete())
self.assertTrue(b.complete())
def test_die_for_non_unique_pending(self):
class A(DummyTask):
def run(self):
logging.debug('running A')
time.sleep(0.1)
super(A, self).run()
a = A()
class B(DummyTask):
def requires(self):
return a
def run(self):
logging.debug('running B')
super(B, self).run()
b = B()
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id='X', keep_alive=True, count_uniques=True) as w:
with Worker(scheduler=sch, worker_id='Y', keep_alive=True, count_uniques=True, wait_interval=0.1) as w2:
self.assertTrue(w.add(b))
self.assertTrue(w2.add(b))
self.assertEqual(w._get_work()[0], 'A()')
self.assertTrue(w2.run())
self.assertFalse(a.complete())
self.assertFalse(b.complete())
def test_complete_exception(self):
"Tests that a task is still scheduled if its sister task crashes in the complete() method"
class A(DummyTask):
def complete(self):
raise Exception("doh")
a = A()
class C(DummyTask):
pass
c = C()
class B(DummyTask):
def requires(self):
return a, c
b = B()
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id="foo") as w:
self.assertFalse(w.add(b))
self.assertTrue(w.run())
self.assertFalse(b.has_run)
self.assertTrue(c.has_run)
self.assertFalse(a.has_run)
def test_requires_exception(self):
class A(DummyTask):
def requires(self):
raise Exception("doh")
a = A()
class D(DummyTask):
pass
d = D()
class C(DummyTask):
def requires(self):
return d
c = C()
class B(DummyTask):
def requires(self):
return c, a
b = B()
sch = CentralPlannerScheduler(retry_delay=100, remove_delay=1000, worker_disconnect_delay=10)
with Worker(scheduler=sch, worker_id="foo") as w:
self.assertFalse(w.add(b))
self.assertTrue(w.run())
self.assertFalse(b.has_run)
self.assertTrue(c.has_run)
self.assertTrue(d.has_run)
self.assertFalse(a.has_run)
