MySQL Monitoring: The Metrics That Actually Matter

Most MySQL Dashboards Track the Wrong Things

The default MySQL monitoring setup tracks hundreds of status variables, most of which are irrelevant to day-to-day operations. Effective monitoring focuses on a small set of metrics that directly indicate problems or predict them before they affect users.

The Essential Five

Track these consistently and you will catch most problems early: InnoDB buffer pool hit ratio (should be above 99%), replication lag (seconds_behind_master plus heartbeat-based measurement), thread running count (active queries, not connections), slow query rate (queries exceeding your SLA threshold), and handler read ratios (indicating full table scans vs index usage).

Alerting Without Noise

The goal is alerts that require action, not dashboards that require interpretation. Set thresholds based on your actual baselines, not generic recommendations. A buffer pool hit ratio of 98% might be fine for one workload and catastrophic for another.

Tools like Percona Monitoring and Management (PMM) provide excellent MySQL-specific visibility out of the box. For existing Prometheus/Grafana stacks, the mysqld_exporter with custom recording rules covers the essentials.