Home / Databases

DatabasesHow-toIntermediate

Database monitoring foundation: the metrics and alerts every DBA team needs

A common monitoring baseline for SQL Server, PostgreSQL, Oracle and MySQL. Covers availability, capacity, performance, replication and backup alerts with thresholds, plus the dashboards and query-level telemetry worth collecting.

Deepak Bhardwaj3 min read

Applies toSQL Server, PostgreSQL, Oracle, MySQL with Prometheus exporters, Azure Monitor, CloudWatch, OCI Monitoring, SolarWinds, Redgate or similar

1. Alert categories and starting thresholds

Category Alert Starting threshold Severity
Availability Instance down / can't connect 1 min Critical
Availability HA replica not synchronising 5 min Critical
Replication Replica or standby lag > the RPO (for example 60 s) Major
Backup No successful log backup 2× the interval Major
Backup No successful full backup 2× the interval Major
Capacity Data or log disk used > 80% warn, > 90% critical Major/Critical
Capacity Transaction log full / WAL piling up Log > 70% used, or WAL directory growing Major
Performance CPU > 85% for 15 min Warning
Performance Blocking Blocked sessions > 60 s Major
Performance Deadlocks Above the baseline Warning
Connections Connections used > 80% of max Major
Errors Severity ≥ 17 errors (SQL Server), FATAL/PANIC (PostgreSQL), ORA-00600/07445 (Oracle) Any Major
Security Failed logins Spike above the baseline Warning

Tune thresholds per tier after 2–4 weeks of baseline data.

2. Collect, don't just alert

  • Query telemetry: Query Store (SQL Server), pg_stat_statements (PostgreSQL), AWR/ASH (Oracle, licensed) or Statspack, Performance Schema plus sys (MySQL).
  • Wait statistics: what the database spends its time waiting on.
  • Trends: database size growth per week, for capacity forecasts.
-- PostgreSQL: top queries by total time
SELECT left(query, 80) AS q, calls, round(total_exec_time) AS total_ms, round(mean_exec_time,1) AS mean_ms
FROM pg_stat_statements ORDER BY total_exec_time DESC LIMIT 10;

3. Tooling options

Stack Tools
Open source Prometheus + postgres_exporter / mysqld_exporter / sql_exporter / oracledb_exporter, Grafana, Alertmanager
Azure Azure Monitor, database watcher for Azure SQL, Log Analytics
AWS CloudWatch, RDS Performance Insights / CloudWatch Database Insights, Enhanced Monitoring
OCI OCI Monitoring, Database Management, Operations Insights
Commercial SolarWinds DPA/SQL Sentry, Redgate Monitor, Datadog DBM, Oracle Enterprise Manager

4. Dashboards

One overview page per tier (all instances, red/amber/green), plus a drill-down per instance: CPU, memory, I/O latency, waits, top queries, connections, replication and backups.

5. Alert hygiene

  • Every alert has an owner and a runbook link.
  • Page people only for Critical alerts. Major alerts become tickets, and Warnings go to dashboards.
  • Monthly review: delete alerts nobody acts on, and add alerts for incidents that weren't caught.

Tip

Alert on missing data too. A monitoring agent that stopped reporting looks exactly like a healthy, quiet database.

Related fixes