def test_calculate_period_end(self):
period = TimePeriod(1000)
start_time = self._get_time1()
period.start_period(start_time)
period_end = period.calculate_period_end()
expected_end_time = start_time + timedelta(seconds=1000)
assert period_end == expected_end_time
def test_calculate_period_end(self):
period = TimePeriod(1000)
start_time = self._get_time1()
period.start_period(start_time)
period_end = period.calculate_period_end()
expected_end_time = start_time + timedelta(seconds=1000)
assert period_end == expected_end_time
def test_calculate_period_end_with_unstarted_period(self):
with pytest.raises(ValueError):
period = TimePeriod(1000)
period_end = period.calculate_period_end()
assert period_end is None
def test_calculate_period_end_with_unstarted_period(self):
with pytest.raises(ValueError):
period = TimePeriod(1000)
period_end = period.calculate_period_end()
assert period_end is None
class ScalingPolicyInstance:
_latest_metric_update = None
state = None
policy = None
instance_uuid = None
threshold_period = None
def __init__(self, instance_uuid, policy):
self._logger = logging.getLogger(__name__)
self.state = ScalingStateUntriggered()
threshold_period_in_seconds = policy.get_threshold_period_in_seconds()
self._logger.debug("threshold period of policy %s is %d", policy, threshold_period_in_seconds)
self.threshold_period = TimePeriod(threshold_period_in_seconds)
self._latest_metric_update = None
self.instance_uuid = instance_uuid
self.policy = policy
def update_policy_state_and_get_scaling_actions(self, metrics_message):
scaling_actions = []
for metric_value in metrics_message.metric_values:
value = metric_value.value
timestamp = dateutil.parser.parse(metric_value.timestamp)
if not self._is_new_metric(timestamp):
self._logger.debug("metric value %s is outdated for instance %s", repr(metric_value), self.instance_uuid)
continue
if self.policy.cooldown_period.is_in_period(timestamp):
self._logger.debug("metric value %s ignored because of cooldown is active", repr(metric_value))
continue
predicate_satisfied = self.policy.value_exceeds_or_undercuts_threshold(float(value))
self._logger.debug("metric value %s is %s under/over threshold",
repr(metric_value),
"" if predicate_satisfied else "not"
)
cooldown_start_time = timestamp
if self.threshold_period.period_started:
cooldown_start_time = self.threshold_period.calculate_period_end()
action_required = self.state.update_and_report_if_action_required(
self,
predicate_satisfied,
timestamp,
metrics_message.uuid
)
if action_required:
scaling_action = self.policy.create_scaling_action(metrics_message.metric_name, self.instance_uuid)
self.policy.cooldown_period.start_period(cooldown_start_time)
scaling_actions.append(scaling_action)
self._logger.info("Triggered scaling action %s, beginning cooldown", repr(scaling_action))
return scaling_actions
def _is_new_metric(self, timestamp):
if self._latest_metric_update is None:
self._latest_metric_update = timestamp
return True
new_metric = timestamp > self._latest_metric_update
if new_metric:
self._latest_metric_update = timestamp
return new_metric
def __repr__(self):
return "ScalingPolicyInstance(instance_uuid={},state={},threshold_period={},policy={})".format(
self.instance_uuid,
repr(self.state),
repr(self.threshold_period),
self.policy.policy_name
)
class ScalingPolicyInstance:
_latest_metric_update = None
state = None
policy = None
instance_uuid = None
threshold_period = None
def __init__(self, instance_uuid, policy):
self._logger = logging.getLogger(__name__)
self.state = ScalingStateUntriggered()
threshold_period_in_seconds = policy.get_threshold_period_in_seconds()
self._logger.debug("threshold period of policy %s is %d", policy,
threshold_period_in_seconds)
self.threshold_period = TimePeriod(threshold_period_in_seconds)
self._latest_metric_update = None
self.instance_uuid = instance_uuid
self.policy = policy
def update_policy_state_and_get_scaling_actions(self, metrics_message):
scaling_actions = []
for metric_value in metrics_message.metric_values:
value = metric_value.value
timestamp = dateutil.parser.parse(metric_value.timestamp)
if not self._is_new_metric(timestamp):
self._logger.debug(
"metric value %s is outdated for instance %s",
repr(metric_value), self.instance_uuid)
continue
if self.policy.cooldown_period.is_in_period(timestamp):
self._logger.debug(
"metric value %s ignored because of cooldown is active",
repr(metric_value))
continue
predicate_satisfied = self.policy.value_exceeds_or_undercuts_threshold(
float(value))
self._logger.debug("metric value %s is %s under/over threshold",
repr(metric_value),
"" if predicate_satisfied else "not")
cooldown_start_time = timestamp
if self.threshold_period.period_started:
cooldown_start_time = self.threshold_period.calculate_period_end(
)
action_required = self.state.update_and_report_if_action_required(
self, predicate_satisfied, timestamp, metrics_message.uuid)
if action_required:
scaling_action = self.policy.create_scaling_action(
metrics_message.metric_name, self.instance_uuid)
self.policy.cooldown_period.start_period(cooldown_start_time)
scaling_actions.append(scaling_action)
self._logger.info(
"Triggered scaling action %s, beginning cooldown",
repr(scaling_action))
return scaling_actions
def _is_new_metric(self, timestamp):
if self._latest_metric_update is None:
self._latest_metric_update = timestamp
return True
new_metric = timestamp > self._latest_metric_update
if new_metric:
self._latest_metric_update = timestamp
return new_metric
def __repr__(self):
return "ScalingPolicyInstance(instance_uuid={},state={},threshold_period={},policy={})".format(
self.instance_uuid, repr(self.state), repr(self.threshold_period),
self.policy.policy_name)
