def test_eval_filter(self):
db_events_filter = EventsSeverityChangerFilter(
new_severity=Severity.NORMAL, event_class=DatabaseLogEvent)
event = DatabaseLogEvent.BAD_ALLOC()
self.assertEqual(event.severity, Severity.ERROR)
db_events_filter.eval_filter(event)
self.assertEqual(event.severity, Severity.NORMAL)
def ignore_alternator_client_errors():
with ExitStack() as stack:
stack.enter_context(
EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=PrometheusAlertManagerEvent,
regex=".*YCSBTooManyErrors.*",
extra_time_to_expiration=60,
))
stack.enter_context(
EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=PrometheusAlertManagerEvent,
regex=".*YCSBTooManyVerifyErrors.*",
extra_time_to_expiration=60,
))
stack.enter_context(
EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=YcsbStressEvent,
regex=r".*Cannot achieve consistency level.*",
extra_time_to_expiration=30,
))
stack.enter_context(
EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=YcsbStressEvent,
regex=r".*Operation timed out.*",
extra_time_to_expiration=30,
))
yield
def ignore_operation_errors():
with ExitStack() as stack:
stack.enter_context(EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=LogEvent,
regex=r".*Operation timed out.*",
extra_time_to_expiration=30,
))
stack.enter_context(EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=LogEvent,
regex=r".*Operation failed for system.paxos.*",
extra_time_to_expiration=30,
))
yield
def test_severity_changer_db_log(self):
"""
See https://github.com/scylladb/scylla-cluster-tests/issues/2115
"""
# 1) Lower DatabaseLogEvent to WARNING for 1 sec.
with self.wait_for_n_events(self.get_events_logger(), count=4, timeout=3):
with EventsSeverityChangerFilter(new_severity=Severity.WARNING,
event_class=DatabaseLogEvent,
extra_time_to_expiration=1):
DatabaseLogEvent.NO_SPACE_ERROR() \
.add_info(node="A", line_number=22, line="critical that should be lowered #1") \
.publish()
DatabaseLogEvent.NO_SPACE_ERROR() \
.add_info(node="A", line_number=22, line="critical that should be lowered #2") \
.publish()
log_content = self.get_event_log_file("warning.log")
self.assertIn("DatabaseLogEvent", log_content)
self.assertIn("critical that should be lowered #1", log_content)
self.assertIn("critical that should be lowered #2", log_content)
# 2) One of the next DatabaseLogEvent event should expire the EventsSeverityChangerFilter
# (and not crash all subscribers)
with self.wait_for_n_events(self.get_events_logger(), count=2, timeout=3):
for _ in range(2):
time.sleep(1)
DatabaseLogEvent.NO_SPACE_ERROR() \
.add_info(node="A", line_number=22, line="critical that shouldn't be lowered") \
.publish()
log_content = self.get_event_log_file("error.log")
self.assertIn("critical that shouldn't be lowered", log_content)
def start_events_device(
log_dir: Optional[Union[str, Path]] = None,
_registry: Optional[EventsProcessesRegistry] = None) -> None:
if _registry is None:
if log_dir is None:
raise RuntimeError(
"Should provide log_dir or instance of EventsProcessesRegistry"
)
_registry = create_default_events_process_registry(log_dir=log_dir)
start_events_main_device(_registry=_registry)
time.sleep(EVENTS_DEVICE_START_DELAY)
start_events_logger(_registry=_registry)
start_grafana_pipeline(_registry=_registry)
start_events_analyzer(_registry=_registry)
time.sleep(EVENTS_SUBSCRIBERS_START_DELAY)
# Default filters.
EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=DatabaseLogEvent.DATABASE_ERROR,
regex=
r'.*workload prioritization - update_service_levels_from_distributed_data: an '
r'error occurred while retrieving configuration').publish()
DbEventsFilter(db_event=DatabaseLogEvent.BACKTRACE,
line='Rate-limit: supressed').publish()
DbEventsFilter(db_event=DatabaseLogEvent.BACKTRACE,
line='Rate-limit: suppressed').publish()
atexit.register(stop_events_device, _registry=_registry)
def test_severity_changer(self):
extra_time_to_expiration = 10
with self.wait_for_n_events(self.get_events_logger(),
count=5,
timeout=3):
with EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=TestFrameworkEvent,
extra_time_to_expiration=extra_time_to_expiration):
TestFrameworkEvent(source="critical that should be lowered #1",
source_method="",
severity=Severity.CRITICAL).publish()
TestFrameworkEvent(source="critical that should be lowered #2",
source_method="",
severity=Severity.CRITICAL).publish()
event = TestFrameworkEvent(
source="critical that should not be lowered #3",
source_method="",
severity=Severity.CRITICAL)
event.source_timestamp = time.time() + extra_time_to_expiration
event.publish()
log_content = self.get_event_log_file("warning.log")
crit_log_content = self.get_event_log_file("critical.log")
self.assertIn("TestFrameworkEvent", log_content)
self.assertIn("critical that should be lowered #1", log_content)
self.assertIn("critical that should be lowered #2", log_content)
self.assertNotIn("critical that should not be lowered #3", log_content)
self.assertIn("critical that should not be lowered #3",
crit_log_content)
def _stop_load_when_nemesis_threads_end(self):
for nemesis_thread in self.db_cluster.nemesis_threads:
nemesis_thread.join()
with EventsSeverityChangerFilter(
new_severity=Severity.
NORMAL, # killing stress creates Critical error
event_class=CassandraStressEvent,
extra_time_to_expiration=60):
self.loaders.kill_cassandra_stress_thread()
def ignore_stream_mutation_fragments_errors():
with ExitStack() as stack:
stack.enter_context(
EventsSeverityChangerFilter(
new_severity=Severity.WARNING,
event_class=DatabaseLogEvent,
regex=r".*Failed to handle STREAM_MUTATION_FRAGMENTS.*",
extra_time_to_expiration=30))
yield
def _stop_load_after_one_nemesis_cycle(self):
time.sleep(300) # wait 5 minutes to be sure nemesis has started
self.db_cluster.stop_nemesis(
timeout=None
) # wait for Nemesis to end and don't start another cycle
with EventsSeverityChangerFilter(
new_severity=Severity.
NORMAL, # killing stress creates Critical error
event_class=CassandraStressEvent,
extra_time_to_expiration=60):
self.loaders.kill_cassandra_stress_thread()