Observability: logs, metrics and traces

In short: Observability means understanding what is happening in a running system from logs, metrics and traces, even for errors you did not foresee.

In more detail: Logs are single events as text. Metrics are numbers over time (requests per second, error rate, memory). Traces follow a request through several services and show where time is lost.

In Depth

• Tools: Prometheus (metrics), Grafana (dashboards), Loki or Elasticsearch (logs), OpenTelemetry (standard for telemetry), Sentry (error reports).
• Golden signals: latency, traffic, errors, saturation.
• Alerts: Alert only on impact for users, not on every trifle.
• Logs done right: structured (JSON), with correlation ID, without passwords and personal data.
• Privacy: Logs with IP addresses are personal data and need retention limits.

See also: CI/CD pipeline, Health check and status page