Monitoring
Overview
Monitoring tells you whether the system is healthy, degraded, or failing.
Why It Matters
Without monitoring, incidents last longer and feel more random than they need to.
Core Concepts
- Metrics show trends.
- Logs show detail.
- Traces show request paths.
Mental Models
If you cannot observe the problem, you cannot confidently fix or verify it.
Best Practices
- Track availability and latency.
- Alert on user-impacting signals.
- Keep dashboards simple enough to use during stress.
Common Mistakes
- Alerting on noise.
- Tracking the wrong metrics.
- Building dashboards no one trusts.
Trade-offs
More monitoring improves visibility, but too much noise weakens the signal.
Decision Framework
| Signal | Use for |
|---|---|
| Metrics | Trends and thresholds |
| Logs | Debugging details |
| Traces | Request path analysis |
Examples
- Alert when error rate rises and latency degrades together.
Checklists
- Can we see the failing path?
- Are alerts actionable?
- Do dashboards answer the common incident questions?
Senior Engineer Notes
Senior engineers want fewer, better signals. Visibility should shorten incidents, not create another dashboard to babysit.