Question

A CloudOps engineer manages an EC2 Auto Scaling group that processes long-running batch jobs. Each instance can take up to two hours to finish a job. During target-tracking scale-in events triggered by low average CPU, the ASG frequently terminates instances that are still mid-job, causing the batch work to be lost and reprocessed. The team wants scale-in to avoid terminating instances that are actively processing a job, while still allowing the group to shrink when instances become idle. What is the MOST appropriate solution?

Accepted Answer

Have the application enable instance scale-in protection on an instance when it starts a job and disable it when the job completes.

Answer

Increase the health check grace period on the Auto Scaling group so instances have more time before termination.

Answer

Switch the scaling policy from target tracking to a step scaling policy based on CPU utilization.

Answer

Set the Auto Scaling group's default cooldown period to two hours to match the maximum job duration.