Use appropriate metric types

When instrumenting applications with metrics, choose the correct metric type based on what you’re measuring to ensure accurate and useful observability data:

1. Use monotonic (increase-only) counters for events or cumulative quantities to ensure compatibility with systems like Prometheus that expect strictly increasing values.

2. Use histograms or distributions for durations rather than simple counters, as durations represent a distribution of values.

3. Consider specialized gauge types for specific use cases, such as high-water mark gauges for tracking peak values:

// GOOD: Using a histogram/distribution for request durations
const requestDuration = createHistogram('api.request.duration.ms');
function handleRequest(req, res) {
  const startTime = performance.now();
  // Process request...
  requestDuration.record(performance.now() - startTime);
}

// GOOD: Using a monotonic counter for counting events
const apiCalls = createCounter('api.calls.total');
function handleRequest(req, res) {
  apiCalls.increment();
  // Process request...
}

// GOOD: Using a high-water mark gauge for tracking peak memory usage
const memoryHighWaterMark = createHighWaterMarkGauge('memory.peak.bytes');
setInterval(() => {
  memoryHighWaterMark.update(process.memoryUsage().heapUsed);
}, 1000);

Choosing the right metric type ensures that monitoring systems can correctly interpret and visualize your data, leading to more meaningful insights and better observability.