Course contentsShow
← Software Engineering
Lesson 1137 of 1,93334. Logging and ObservabilityFree lesson

The Four Golden Signals

Monitoring latency, traffic, errors, and saturation as core health indicators for any system.

The Four Golden Signals

What you'll learn: How to monitor any system's health using four essential metrics that reveal most problems before they become critical.

The Core Health Indicators

Imagine a doctor checking your vital signs—heart rate, blood pressure, temperature, and breathing. Similarly, Google's Site Reliability Engineering team identified four golden signals that tell you almost everything about your system's health:

1. Latency

How long requests take to complete. This includes both successful requests and failures (which often complete faster but shouldn't be ignored). Rising latency means users are waiting longer, even if requests succeed.

2. Traffic

How much demand your system is handling—requests per second, transactions per minute, or data throughput. This measures workload intensity and helps identify usage patterns or attacks.

3. Errors

The rate of failed requests, whether explicit (HTTP 500s) or implicit (returning wrong data with HTTP 200). High error rates mean users are experiencing failures, even if the system is technically "up."

4. Saturation

How "full" your service is—CPU, memory, disk I/O, network capacity. When resources approach limits, performance degrades. This is your early warning system before complete failure.

Why These Four?

These signals complement each other perfectly. Metrics (covered previously) give you the measurements, but these four tell you which metrics matter most. They work together: high traffic might cause saturation, leading to increased latency and eventually errors.

Unlike logs (event records) or traces (request flows), these golden signals provide the 10,000-foot view. They answer: "Is my service healthy right now?"

Key Takeaway: Monitor latency, traffic, errors, and saturation as your primary system health indicators—together they catch the vast majority of issues before users complain.