Observability

Observability means being able to understand what a running system is doing from the information it produces.

Metrics

Useful backend metrics include requests per second, error rate, p50 latency, p95 latency, p99 latency, CPU, memory and queue depth.

A practical model is:

Rate
Errors
Duration

Structured Logs

{
  "level": "error",
  "event": "payment_capture_failed",
  "request_id": "req_123",
  "order_id": "ord_77"
}

Never casually log passwords, access tokens or secrets.

Distributed Tracing

Correlation IDs

Attach a request ID or trace ID to related logs.

Useful observability reduces the time between noticing a problem and understanding its cause.