October 25, 2013

Amazon EC2 Auto Scaling example

The following code creates an Auto Scaling group with scale-up and scale-down policies based on the monitoring the CPU utilization of either a specific EC2 instance or instances within the Auto Scaling group cluster.

If alarm_dimensions specifies InstanceId, the Auto Scaling group will scale up or down according to the CPU utilization of the specified running instance; if alarm_dimensions specifies AutoScalingGroupName, the Auto Scaling group will scale up or down according to the CPU utilization of instances within its existing cluster. For more metrics and dimensions in CloudWatch, see Amazon Elastic Compute Cloud Dimensions and Metrics.

Raw code can be downloaded here.

import boto.ec2
from boto.ec2.autoscale import (AutoScalingGroup,
                                LaunchConfiguration, ScalingPolicy)
from boto.ec2.cloudwatch import MetricAlarm


# Create connections to Auto Scaling and CloudWatch
as_conn = boto.ec2.autoscale.connect_to_region("us-east-1")
cw_conn = boto.ec2.cloudwatch.connect_to_region("us-east-1")

# Name for auto scaling group and launch configuration
as_name = "VM1"

# Create launch configuration
lc = LaunchConfiguration(name=as_name,
                         image_id="ami-76f0061f",  # AMI ID of your instance
                         key_name="your_key_name",
                         security_groups="your_group_name",
                         instance_type="t1.micro",
                         instance_monitoring=True)

as_conn.create_launch_configuration(lc)

# Create Auto Scaling group
ag = AutoScalingGroup(group_name=as_name,
                      availability_zones=["us-east-1b"],
                      launch_config=lc, min_size=0,
                      max_size=2,
                      connection=as_conn)

as_conn.create_auto_scaling_group(ag)

# Create scaling policies
scale_up_policy = ScalingPolicy(
    name='scale_up', adjustment_type='ChangeInCapacity',
    as_name=as_name, scaling_adjustment=1, cooldown=180)

scale_down_policy = ScalingPolicy(
    name='scale_down', adjustment_type='ChangeInCapacity',
    as_name=as_name, scaling_adjustment=-1, cooldown=180)

as_conn.create_scaling_policy(scale_up_policy)
as_conn.create_scaling_policy(scale_down_policy)

scale_up_policy = as_conn.get_all_policies(
    as_group=as_name, policy_names=['scale_up'])[0]

scale_down_policy = as_conn.get_all_policies(
    as_group=as_name, policy_names=['scale_down'])[0]

# Set dimensions for CloudWatch alarms
# Monitor on a specific instance
alarm_dimensions = {"InstanceId": "your_instance_id"}

# Monitor instances within the Auto Scaling group cluster
alarm_dimensions_as = {"AutoScalingGroupName": as_name}

# Create metric alarms
scale_up_alarm = MetricAlarm(
    name='scale_up_on_cpu_' + as_name, namespace='AWS/EC2',
    metric='CPUUtilization', statistic='Average',
    comparison='>', threshold="80",
    period='60', evaluation_periods=2,
    alarm_actions=[scale_up_policy.policy_arn],
    dimensions=alarm_dimensions)

scale_down_alarm = MetricAlarm(
    name='scale_down_on_cpu_' + as_name, namespace='AWS/EC2',
    metric='CPUUtilization', statistic='Average',
    comparison='<', threshold="20",
    period='60', evaluation_periods=2,
    alarm_actions=[scale_down_policy.policy_arn],
    dimensions=alarm_dimensions)

# Create alarms in CloudWatch
cw_conn.create_alarm(scale_up_alarm)
cw_conn.create_alarm(scale_down_alarm)
Posted by Shichao An
Tags: Python, AWS