Ops wants to alarm on error RATE (5xx ÷ total requests), not raw error count — CloudWatch has no such metric natively. Which feature computes it?
A.CloudWatch metric math — an expression combining existing metrics
Correct. Expressions over existing metrics, alarmable.
B.A custom application metric published per request
Works but requires code changes — metric math doesn't.
C.Metric filters on the load balancer logs
Filters extract counts from logs, not live ratios.
D.CloudWatch anomaly detection
Anomaly detection bands metrics — doesn't combine them.
Why this answer
Metric math lets you compute ratios/aggregates across metrics (e.g., m2/m1*100) and alarm on the result — no new instrumentation needed. Publishing a custom metric works but adds app changes and cost.
Correct answer: A · Last reviewed 2026-11-03
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Monitoring & Remediation practice session.
Practice Monitoring & Remediation