Quick Notes85 words
Metrics and queries for operations — quick notes
Operations metrics — quick notes
| Layer | Use for |
|---|---|
| User-facing — availability, latency, error rate | Alerting |
| Infrastructure — CPU, memory, disk, network | Diagnosis |
- Report p95 / p99, not averages — averages hide the tail.
- Azure Monitor logs are queried with KQL: filter → summarise → order.
- An alert must name something a human can do.
Trap: alerting on infrastructure symptoms creates alert fatigue, and fatigue is why a real page gets ignored.