A company runs custom network-analysis software to inspect traffic entering and leaving a VPC. The software is deployed with an AWS CloudFormation template onto three Amazon EC2 instances that are managed by an Auto Scaling group. Network routing is configured so traffic is directed to the EC2 instances. When the analysis software stops working, the Auto Scaling group replaces the failed instance, but the network routes are not updated to reference the replacement instance. Which combination of actions will solve this problem? (Choose three.)

The problem is that the Auto Scaling group replaces a failed instance, but the network routes aren't updated. To solve this, first, the CloudWatch agent should be installed via the CloudFormation template to send process-level metrics for the custom application. This allows monitoring the application's health, not just the EC2 instance's. Second, a CloudWatch alarm should be created for a custom application metric that indicates failure, publishing to an SNS topic. This triggers an alert when the application itself fails. Finally, an AWS Lambda function, triggered by the SNS notification, can then take the failed instance out of service and, critically, update the network routes to point to the new, healthy instance. Incorrect options: Creating alarms based on EC2 status checks only detects instance-level failures, not application-level failures. Installing AWS Systems Manager Agent for process metrics is possible but CloudWatch Agent is the standard for sending custom metrics to CloudWatch. Adding a condition in CloudFormation for automatic route updates on launch is insufficient because the route update needs to happen dynamically when an instance is replaced due to failure, not just on initial launch.