How to automate an EC2 reboot after a failed instance health check?

Why the Answer Is Correct

StatusCheckFailed_Instance reports failures that originate inside the guest OS, such as a hung kernel, exhausted memory, or a corrupted network stack, and the standard remediation for that class of failure is a reboot. Option C builds the alarm on exactly that metric, uses the instance ID as the dimension so the alarm targets only this non-production instance, and configures an EC2 reboot action when the value is greater than zero. That precisely mirrors the manual workflow described in the scenario. No EventBridge service-linked role is actually needed for a CloudWatch alarm-to-EC2 action, but its presence in the stem is a distractor rather than a blocker.

Why the Other Options Are Wrong

Option A uses HealthyHostCount, which is an Elastic Load Balancing metric, not an EC2 status check, and firing a reboot when the count is greater than zero would reboot a healthy target. Option B selects the correct metric but pairs it with the recover action, which is intended for StatusCheckFailed_System host-level failures and cannot repair an instance-level OS problem. Option D uses the AMI ID as the dimension, which will never match a StatusCheckFailed_Instance datapoint, disables the benefit of detailed monitoring (status checks are always published at one-minute granularity), and performs a stop/start, which is a different operation from the required reboot.

Community Comment Notes

As Bad_Mat noted, StatusCheckFailed_Instance "tracks failures inside the instance itself" such as OS or software/configuration problems, which is why a reboot is the appropriate remediation. The same comment stresses that the alarm action should mirror the engineer's manual reboot, and that the instance ID dimension scopes the automation to the single non-production instance. That reasoning aligns with the vendor's documented guidance for instance status-check alarms.