如何为Auto Scaling组中的每个Windows实例创建独立的CloudWatch磁盘存储告警，并在实例缩容时自动删除告警？

Short Answer #

Absolutely, this is fully feasible! You can create individual CloudWatch metrics and alarms for every instance in your Auto Scaling Group (ASG), and automate cleanup of these alarms when instances are terminated during scaling-in operations.

Step-by-Step Implementation #

1. Ensure Per-Instance Disk Metrics Are Being Collected #

Windows instances don’t send detailed disk metrics (like % Free Space or Free Megabytes) to CloudWatch by default. You’ll need to set up the CloudWatch Agent on each instance:

• Add a script to your ASG’s Launch Template/User Data to install the CloudWatch Agent automatically when a new instance spins up. For example:

  msiexec.exe /i https://s3.amazonaws.com/amazoncloudwatch-agent/windows/amd64/latest/amazon-cloudwatch-agent.msi /qn
  

• Create a CloudWatch Agent configuration file (store it in SSM Parameter Store for easy access) that specifies disk metrics and includes the InstanceId as a dimension. Example snippet from the config:

  {
    "metrics": {
      "metrics_collected": {
        "LogicalDisk": {
          "measurement": [
            "% Free Space",
            "Free Megabytes"
          ],
          "dimensions": [{"InstanceId": "${aws:InstanceId}"}],
          "resources": ["*"]
        }
      }
    }
  }
  

• Use the agent to load this config so metrics are sent to CloudWatch with the instance-specific dimension.

2. Automate Per-Instance Alarm Creation #

Skip manual alarm setup by automating creation when new instances launch:

• Option 1: Lambda + ASG Lifecycle Hooks
  • Set up an ASG lifecycle hook for the EC2_INSTANCE_LAUNCH_SUCCESS event. This triggers a Lambda function when a new instance is ready.
  • The Lambda function grabs the instance ID from the event, then calls the CloudWatch API (PutMetricAlarm) to create a disk-specific alarm for that instance. Use a consistent naming convention like Disk-Free-Space-Alarm-{InstanceId} to simplify later cleanup.
• Option 2: CloudFormation/Terraform
  • Define an alarm resource that references the instance ID as a parameter. Trigger a stack creation for each new instance using automation tools tied to ASG launch events.

3. Automate Alarm Cleanup During Scaling-In #

To delete alarms when instances are terminated:

• Option 1: Lambda + ASG Lifecycle Hooks
  • Set up another lifecycle hook for the EC2_INSTANCE_TERMINATING event. When an instance is marked for termination, trigger a Lambda function.
  • The Lambda function uses the instance ID from the event to find all alarms matching your naming convention (e.g., Disk-Free-Space-Alarm-i-0123456789abcdef0) and calls the CloudWatch API (DeleteAlarms) to remove them. Don’t forget to complete the lifecycle hook so the instance can terminate properly.
• Option 2: EventBridge + Lambda
  • Create an EventBridge rule that captures EC2 Instance Terminate events. Route these events to a Lambda function that cleans up the corresponding alarms.

Key Tips #

• IAM Permissions: Make sure your Lambda role has the necessary permissions: cloudwatch:PutMetricAlarm, cloudwatch:DeleteAlarms, autoscaling:CompleteLifecycleAction, and ec2:DescribeInstances.
• Alarm Thresholds: Define thresholds that fit your workload (e.g., alert when % Free Space drops below 10%).
• Testing: Test the workflow manually first—launch an instance to verify the alarm is created, then terminate it to confirm the alarm is deleted.

内容的提问来源于stack exchange，提问作者Lakshay Kashyap