Which of these solutions would meet the requirements?
Create an Amazon CloudWatch alarm on the CPUCreditBalance metric and specify a terminate alarm action.
Create an Amazon CloudWatch alarm on the CPUUtilization metric and specify a reboot alarm action.
Create an Amazon CloudWatch alarm on the CPUCreditBalance metric and specify a reboot alarm action.
Create an Amazon CloudWatch alarm on the CPUUtilization metric and specify a terminate alarm action.
Explanations:
The CPUCreditBalance metric measures burstable performance instances’ remaining CPU credits, not CPU utilization. A terminate action would not solve the requirement of rebooting instances based on high CPU usage.
The CPUUtilization metric tracks the percentage of CPU usage. A CloudWatch alarm with a reboot action would automatically reboot instances when their CPU usage exceeds 80%, fulfilling the requirement.
The CPUCreditBalance metric relates to burstable instances and does not track CPU utilization. A reboot action on this metric would not address the requirement to reboot based on CPU usage.
The CPUUtilization metric does track CPU usage, but a terminate alarm action would terminate the instance, not reboot it. This does not meet the requirement of rebooting instances based on high CPU usage.
In my experience, the answer is:
Create an Amazon CloudWatch alarm on the CPUUtilization metric and specify a reboot alarm action.