Symptom-based alerting focuses on monitoring signals that directly measure the user-facing impact or the health of a service from an external perspective. High API latency is a direct symptom of a degraded user experience, indicating that the service is slow for its consumers. This approach prioritizes alerts that are immediately actionable and relevant to service-level objectives (SLOs), rather than focusing on underlying component metrics which may not correlate with a real problem.

B. Disk space: This is a cause-based metric. Low disk space is a potential cause of a future failure but is not, in itself, a symptom that users are currently experiencing.

C. Database availability: This is a cause-based alert. The database being unavailable is the cause of a problem; the resulting symptom would be application errors or high latency.

D. CPU usage: This is a classic cause-based metric. High CPU utilization is an internal system state that may or may not lead to a user-visible symptom like slowness.

---

1. Google SRE Book. Beyer, B., Jones, C., Petoff, J., & Murphy, N. R. (2016). Site Reliability Engineering: How Google Runs Production Systems. O'Reilly Media.

Chapter 6, "Monitoring Distributed Systems," Section: "Symptoms Versus Causes." This section explicitly distinguishes between symptoms (what's broken from a user perspective) and causes (why it's broken). It states, "For example, if the symptom is 'latency is high,' the cause might be that your database is slow." This directly frames latency as a symptom.

2. Carnegie Mellon University, School of Computer Science. (Fall 2023). 15-319/619 Cloud Computing Courseware.

Lecture 20: "Monitoring and SRE," Slide 13: "Symptom-based vs. Cause-based Monitoring." The slide categorizes "Latency" and "Error Rate" as examples of symptom-based monitoring, while "CPU Load" and "Disk Space" are listed as examples of cause-based monitoring.

3. Prometheus Documentation. Best Practices: Alerting.

Section: "Alerting." The official documentation for Prometheus, a core tool in modern Linux environments, advises: "Alerting rules should be configured in Prometheus to fire on symptoms that are visible to the users of a service... For example, the alert HighRequestLatency is a symptom-based alert because it is directly visible to the users of the API."