SOA-C02 Practice Question 1: CloudWatch Agent Memory Metrics
AWS Certified SysOps Administrator – Associate (SOA-C02) · Test 1 · question 1 of 10
An operations team builds a CloudWatch dashboard for a fleet of Amazon EC2 Linux instances. The dashboard shows CPUUtilization, NetworkIn, and DiskReadBytes, but the team cannot find any metric for memory utilization, so they cannot create the memory alarm their runbook requires.
What should the team do to make memory utilization available in CloudWatch?
Show answer and explanation
Answer:
EC2 publishes only hypervisor-visible metrics by default. Memory utilization, swap usage, and filesystem-level disk space live inside the guest operating system, so AWS cannot see them. The unified CloudWatch agent runs inside the instance, collects mem_used_percent, disk_used_percent and similar values, and publishes them to CloudWatch as custom metrics that dashboards and alarms can use.
Option B is wrong because detailed monitoring changes only the *frequency* of the existing EC2 metrics, from 5-minute to 1-minute periods. It does not add any new metric, and memory is still absent.
Option C is wrong because metric math combines metrics that already exist; there is no arithmetic relationship between CPU, network throughput, and memory consumption that could produce a valid memory figure.
Option D is wrong because instance and system status checks report only reachability and underlying-hardware impairment as pass/fail values. They expose nothing about guest memory.
Reference: Metrics collected by the CloudWatch agent