An EC2 application is failing from memory pressure, but standard EC2 monitoring shows only low CPU. Operations needs guest memory and filesystem utilization. Which approach is appropriate?

1. Use CloudTrail management events as a continuous memory-percentage metric.
2. Enable EC2 detailed monitoring and assume it adds every guest operating-system metric.
3. Infer free memory solely from the number of security group rules.
4. Configure the CloudWatch agent for the required guest metrics and grant the necessary publishing permissions.