def __call__(self, parser, namespace, values, option_string=None):
from azure.mgmt.monitor.models import MetricTrigger
if len(values) == 1:
# workaround because CMD.exe eats > character... Allows condition to be
# specified as a quoted expression
values = values[0].split(' ')
name_offset = 0
try:
metric_name = ' '.join(values[name_offset:-4])
operator = get_autoscale_operator_map()[values[-4]]
threshold = int(values[-3])
aggregation = get_autoscale_aggregation_map()[values[-2].lower()]
window = period_type(values[-1])
except (IndexError, KeyError):
from knack.util import CLIError
raise CLIError('usage error: --condition METRIC {==,!=,>,>=,<,<=} '
'THRESHOLD {avg,min,max,total,count} PERIOD')
condition = MetricTrigger(
metric_name=metric_name,
metric_resource_uri=None, # will be filled in later
time_grain=None, # will be filled in later
statistic=None, # will be filled in later
time_window=window,
time_aggregation=aggregation,
operator=operator,
threshold=threshold)
namespace.condition = condition
def validate_autoscale_timegrain(namespace):
from azure.mgmt.monitor.models import MetricTrigger
from azure.cli.command_modules.monitor.actions import period_type
from azure.cli.command_modules.monitor.util import get_autoscale_statistic_map
values = namespace.timegrain
if len(values) == 1:
# workaround because CMD.exe eats > character... Allows condition to be
# specified as a quoted expression
values = values[0].split(' ')
name_offset = 0
try:
time_grain = period_type(values[1])
name_offset += 1
except ValueError:
time_grain = period_type('1m')
try:
statistic = get_autoscale_statistic_map()[values[0]]
name_offset += 1
except KeyError:
statistic = get_autoscale_statistic_map()['avg']
timegrain = MetricTrigger(metric_name=None,
metric_resource_uri=None,
time_grain=time_grain,
statistic=statistic,
time_window=None,
time_aggregation=None,
operator=None,
threshold=None)
namespace.timegrain = timegrain
def create_rule_instance(params):
rule = params.copy()
rule['metric_resource_uri'] = rule.get('metric_resource_uri', self.target)
rule['time_grain'] = timedelta(minutes=rule.get('time_grain', 0))
rule['time_window'] = timedelta(minutes=rule.get('time_window', 0))
rule['cooldown'] = timedelta(minutes=rule.get('cooldown', 0))
return ScaleRule(metric_trigger=MetricTrigger(**rule), scale_action=ScaleAction(**rule))
def result(self):
from azure.mgmt.monitor.models import MetricTrigger, ScaleRuleMetricDimension
dim_params = self.parameters.get('dimensions', [])
dimensions = []
for dim in dim_params:
dimensions.append(ScaleRuleMetricDimension(**dim))
self.parameters['dimensions'] = dimensions
self.parameters['metric_resource_uri'] = None # will be filled in later
self.parameters['time_grain'] = None # will be filled in later
self.parameters['statistic'] = None # will be filled in later
return MetricTrigger(**self.parameters)
def create_autoscaling_settings(ExistingVmScaleSetName, VmScaleSetID,
resourceGroupName):
NewVmScaleSetName = VmScaleSetID.split('/')[-1]
existing_asg = monitor_client.autoscale_settings.get(
resource_group_name=resourceGroupName,
autoscale_setting_name=ExistingVmScaleSetName)
rules = [
ScaleRule(
metric_trigger=MetricTrigger(
metric_name=i.metric_trigger.metric_name,
#metric_namespace=i.metric_trigger.additional_properties['metricNamespace'],
metric_resource_uri=VmScaleSetID,
time_grain=i.metric_trigger.time_grain,
statistic=i.metric_trigger.statistic,
time_window=i.metric_trigger.time_window,
time_aggregation=i.metric_trigger.time_aggregation,
operator=i.metric_trigger.operator,
threshold=i.metric_trigger.threshold
#dimensions = i.metric_trigger.additional_properties['dimensions']
),
scale_action=i.scale_action)
for i in existing_asg.profiles[0].rules
]
profile = AutoscaleProfile(name=existing_asg.profiles[0].name,
capacity=existing_asg.profiles[0].capacity,
rules=rules,
fixed_date=None,
recurrence=None)
parameters = AutoscaleSettingResource(
location=existing_asg.location,
tags=existing_asg.tags,
profiles=[profile],
notifications=existing_asg.notifications,
enabled=True,
autoscale_setting_resource_name=NewVmScaleSetName,
target_resource_uri=VmScaleSetID)
new_asg = monitor_client.autoscale_settings.create_or_update(
resource_group_name=resourceGroupName,
autoscale_setting_name=NewVmScaleSetName,
parameters=parameters)
def shutdown_scaleset_rule(queue_uri: str) -> ScaleRule:
return ScaleRule(
# Scale in if there are 0 or more messages in the queue (aka: every time)
metric_trigger=MetricTrigger(
metric_name="ApproximateMessageCount",
metric_resource_uri=queue_uri,
# Check every 10 minutes
time_grain=timedelta(minutes=5),
# The average amount of messages there are in the pool queue
time_aggregation=TimeAggregationType.AVERAGE,
statistic=MetricStatisticType.SUM,
# Over the past 10 minutes
time_window=timedelta(minutes=5),
operator=ComparisonOperationType.GREATER_THAN_OR_EQUAL,
threshold=0,
divide_per_instance=False,
),
scale_action=ScaleAction(
direction=ScaleDirection.DECREASE,
type=ScaleType.CHANGE_COUNT,
value=1,
cooldown=timedelta(minutes=5),
),
)
def create_auto_scale_profile(
queue_uri: str,
min: int,
max: int,
default: int,
scale_out_amount: int,
scale_out_cooldown: int,
scale_in_amount: int,
scale_in_cooldown: int,
) -> AutoscaleProfile:
return AutoscaleProfile(
name=str(uuid.uuid4()),
capacity=ScaleCapacity(minimum=min, maximum=max, default=max),
# Auto scale tuning guidance:
# https://docs.microsoft.com/en-us/azure/architecture/best-practices/auto-scaling
rules=[
ScaleRule(
metric_trigger=MetricTrigger(
metric_name="ApproximateMessageCount",
metric_resource_uri=queue_uri,
# Check every 15 minutes
time_grain=timedelta(minutes=15),
# The average amount of messages there are in the pool queue
time_aggregation=TimeAggregationType.AVERAGE,
statistic=MetricStatisticType.COUNT,
# Over the past 15 minutes
time_window=timedelta(minutes=15),
# When there's more than 1 message in the pool queue
operator=ComparisonOperationType.GREATER_THAN_OR_EQUAL,
threshold=1,
divide_per_instance=False,
),
scale_action=ScaleAction(
direction=ScaleDirection.INCREASE,
type=ScaleType.CHANGE_COUNT,
value=scale_out_amount,
cooldown=timedelta(minutes=scale_out_cooldown),
),
),
# Scale in
ScaleRule(
# Scale in if no work in the past 20 mins
metric_trigger=MetricTrigger(
metric_name="ApproximateMessageCount",
metric_resource_uri=queue_uri,
# Check every 10 minutes
time_grain=timedelta(minutes=10),
# The average amount of messages there are in the pool queue
time_aggregation=TimeAggregationType.AVERAGE,
statistic=MetricStatisticType.SUM,
# Over the past 10 minutes
time_window=timedelta(minutes=10),
# When there's no messages in the pool queue
operator=ComparisonOperationType.EQUALS,
threshold=0,
divide_per_instance=False,
),
scale_action=ScaleAction(
direction=ScaleDirection.DECREASE,
type=ScaleType.CHANGE_COUNT,
value=scale_in_amount,
cooldown=timedelta(minutes=scale_in_cooldown),
),
),
],
)
