def test_free_cpu_on_container_die(self):
workload_name = str(uuid.uuid4())
workload = Workload(workload_name, DEFAULT_CPU_COUNT, STATIC)
docker_client = MockDockerClient([MockContainer(workload)])
events = [
get_container_create_event(DEFAULT_CPU_COUNT, STATIC,
workload_name, workload_name),
get_container_die_event(workload_name)
]
event_count = len(events)
event_iterable = MockEventProvider(
events, 1) # Force in order event processing for the test
test_context = TestContext(docker_client)
manager = EventManager(event_iterable,
test_context.get_event_handlers(),
get_mock_file_manager(),
DEFAULT_TEST_EVENT_TIMEOUT_SECS)
wait_until(lambda: event_count == manager.get_processed_count())
self.assertEqual(0, manager.get_queue_depth())
self.assertEqual(DEFAULT_TOTAL_THREAD_COUNT,
len(test_context.get_cpu().get_empty_threads()))
self.assertEqual(
1,
test_context.get_create_event_handler().get_handled_event_count())
self.assertEqual(
1,
test_context.get_free_event_handler().get_handled_event_count())
manager.stop_processing_events()
def test_add_metrics(self):
test_context = TestContext()
workload_name = str(uuid.uuid4())
events = [
get_container_create_event(DEFAULT_CPU_COUNT, STATIC,
workload_name, workload_name)
]
event_count = len(events)
event_manager = EventManager(MockEventProvider(events),
test_context.get_event_handlers(),
get_mock_file_manager(), 5.0)
wait_until(lambda: event_count == event_manager.get_processed_count())
log.info("Event manager has processed {} events.".format(
event_manager.get_processed_count()))
workload_manager = test_context.get_workload_manager()
registry = Registry()
reporter = InternalMetricsReporter(workload_manager, event_manager)
reporter.set_registry(registry)
reporter.report_metrics({})
wait_until(lambda: self.__gauge_value_equals(registry, RUNNING, 1))
wait_until(lambda: self.__gauge_value_equals(registry, ADDED_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(registry, REMOVED_KEY, 0))
wait_until(
lambda: self.__gauge_value_equals(registry, SUCCEEDED_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(registry, FAILED_KEY, 0))
wait_until(
lambda: self.__gauge_value_equals(registry, QUEUE_DEPTH_KEY, 0))
wait_until(
lambda: self.__gauge_value_equals(registry, WORKLOAD_COUNT_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(
registry, PACKAGE_VIOLATIONS_KEY, 0))
wait_until(lambda: self.__gauge_value_equals(registry,
CORE_VIOLATIONS_KEY, 0))
wait_until(lambda: self.__gauge_value_equals(registry,
EVENT_SUCCEEDED_KEY, 3))
wait_until(
lambda: self.__gauge_value_equals(registry, EVENT_FAILED_KEY, 0))
wait_until(lambda: self.__gauge_value_equals(registry,
EVENT_PROCESSED_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(
registry, FALLBACK_ALLOCATOR_COUNT, 0))
wait_until(lambda: self.__gauge_value_equals(
registry, IP_ALLOCATOR_TIMEBOUND_COUNT, 0))
event_manager.stop_processing_events()
def test_unknown_workload_type_label(self):
test_context = TestContext()
unknown_event = get_event(
CONTAINER, CREATE, uuid.uuid4(), {
NAME: "container-name",
CPU_LABEL_KEY: "1",
WORKLOAD_TYPE_LABEL_KEY: "unknown"
})
valid_event = get_container_create_event(1)
event_iterable = MockEventProvider([unknown_event, valid_event])
manager = EventManager(event_iterable,
test_context.get_event_handlers(),
get_mock_file_manager(),
DEFAULT_TEST_EVENT_TIMEOUT_SECS)
wait_until(lambda: manager.get_error_count() == 1)
wait_until(lambda: manager.get_processed_count() == 2)
self.assertEqual(0, manager.get_queue_depth())
manager.stop_processing_events()
def test_crash_ip_allocator_metrics(self):
cpu = get_cpu(2, 16, 2)
test_context = TestContext(cpu=cpu)
# now override the cpu seen by the allocator to crash it
test_context.get_workload_manager().get_allocator().set_cpu(
get_cpu(2, 2, 2))
events = [get_container_create_event(10, name="foo", id="bar")]
event_count = len(events)
event_manager = EventManager(MockEventProvider(events),
test_context.get_event_handlers(),
get_mock_file_manager(), 5.0)
wait_until(lambda: event_count == event_manager.get_processed_count())
log.info("Event manager has processed {} events.".format(
event_manager.get_processed_count()))
workload_manager = test_context.get_workload_manager()
registry = Registry()
reporter = InternalMetricsReporter(workload_manager, event_manager)
reporter.set_registry(registry)
reporter.report_metrics({})
wait_until(lambda: self.__gauge_value_equals(registry, RUNNING, 1))
wait_until(lambda: self.__gauge_value_equals(registry, ADDED_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(registry, REMOVED_KEY, 0))
wait_until(
lambda: self.__gauge_value_equals(registry, SUCCEEDED_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(registry, FAILED_KEY, 0))
wait_until(
lambda: self.__gauge_value_equals(registry, WORKLOAD_COUNT_KEY, 1))
wait_until(lambda: self.__gauge_value_equals(
registry, FALLBACK_ALLOCATOR_COUNT, 1))
event_manager.stop_processing_events()
def test_edge_case_ip_allocator_metrics(self):
# this is a specific scenario causing troubles to the solver.
# we should hit the time-bound limit and report it.
cpu = get_cpu(2, 16, 2)
test_context = TestContext(cpu=cpu)
test_context.get_workload_manager().get_allocator(
).set_solver_max_runtime_secs(0.01)
events = []
cnt_evts = 0
for i in range(15):
events.append(get_container_create_event(2, name=str(i),
id=str(i)))
cnt_evts += 15
events.append(get_container_create_event(1, name="15", id="15"))
cnt_evts += 1
for i in range(9):
events.append(
get_container_create_event(2,
name=str(i + cnt_evts),
id=str(i + cnt_evts)))
events.append(get_container_die_event(name="15", id="15"))
event_count = len(events)
event_manager = EventManager(MockEventProvider(events),
test_context.get_event_handlers(),
get_mock_file_manager(), 5.0)
wait_until(lambda: event_count == event_manager.get_processed_count(),
timeout=20)
log.info("Event manager has processed {} events.".format(
event_manager.get_processed_count()))
workload_manager = test_context.get_workload_manager()
registry = Registry()
reporter = InternalMetricsReporter(workload_manager, event_manager)
reporter.set_registry(registry)
reporter.report_metrics({})
wait_until(lambda: self.__gauge_value_equals(registry, RUNNING, 1))
wait_until(lambda: self.__gauge_value_equals(registry, ADDED_KEY, 25))
wait_until(lambda: self.__gauge_value_equals(registry, REMOVED_KEY, 1))
wait_until(
lambda: self.__gauge_value_equals(registry, SUCCEEDED_KEY, 26))
wait_until(lambda: self.__gauge_value_equals(registry, FAILED_KEY, 0))
wait_until(
lambda: self.__gauge_value_equals(registry, QUEUE_DEPTH_KEY, 0))
wait_until(lambda: self.__gauge_value_equals(registry,
WORKLOAD_COUNT_KEY, 24))
wait_until(lambda: self.__gauge_value_equals(
registry, PACKAGE_VIOLATIONS_KEY, 0))
wait_until(lambda: self.__gauge_value_equals(
registry, EVENT_SUCCEEDED_KEY, 3 * 26))
wait_until(
lambda: self.__gauge_value_equals(registry, EVENT_FAILED_KEY, 0))
wait_until(lambda: self.__gauge_value_equals(registry,
EVENT_PROCESSED_KEY, 26))
wait_until(lambda: self.__gauge_value_reached(
registry, IP_ALLOCATOR_TIMEBOUND_COUNT, 1))
wait_until(lambda: self.__gauge_value_reached(
registry, ALLOCATOR_CALL_DURATION, 0.1))
event_manager.stop_processing_events()