Database Monitoring: The Five Numbers That Matter

Monitor too few and you miss issues; too many and they drown in noise. Five is the right number.

Why five

One metric misses cases. Twenty metrics drown signal.

Five-metric monitoring catches 90% of operational problems.

The five metrics

• 1. Connection count vs max.
• 2. Slow query rate.
• 3. Replication lag.
• 4. Lock wait time.
• 5. Disk usage growth.

Dashboard pattern

One panel per metric; trend over 24h; threshold lines visible.

Single dashboard URL; bookmarked by on-call; the database health view.

Alert thresholds

Connection count: alert at 80% of max.

Slow query rate: alert if doubled vs baseline.

Replication lag: alert at 30s.

Lock wait: alert at 5s p99.

Disk growth: alert at 7-day projection > 80%.

Antipatterns

• Monitoring without baseline. Cannot detect anomaly.
• Twenty metrics. Drown signal.
• No alerts on five. Discovery via incident.

What to do this week

Three moves. (1) Apply this pattern to your most-loaded table. (2) Measure query latency / write throughput before/after. (3) Document the win and the constraint so the next refactor inherits the knowledge.