Most Magento stores are flying blind. They know the site is down when customers stop ordering, not when the first error appears. A proper observability stack gives you the signals to catch problems before customers do. Here's what we monitor and how.
Critical Metrics to Monitor
- Order rate per minute — sudden drops indicate checkout problems
- PHP-FPM queue depth — approaching the limit means you need more capacity
- MySQL query time p99 — slow queries impact all concurrent requests
- Redis memory usage — approaching maxmemory triggers evictions
- Varnish cache hit rate — dropping below 85% means more load on PHP
- Magento cron execution time — stuck cron jobs delay email and order processing
- Error rate from exception.log — spikes indicate new bugs or infrastructure problems
Log Aggregation
Magento writes logs to var/log/. Forward these to a centralized log platform (ELK stack, Datadog Logs, or Grafana Loki) using a log agent (Filebeat, Fluentd). Structure your log parsing to extract severity, exception class, and store context. Build a dashboard showing error rate over time with links to log entries for quick debugging.
Alerting Strategy
Alert on symptoms, not causes. Page your on-call engineer when the order rate drops 50% below baseline for 5 minutes — don't alert on individual PHP errors (too noisy). Create tiered alerts: PagerDuty for revenue-impacting issues, Slack for anomalies that need investigation, email digest for trends. Every alert must have a runbook link — your on-call engineer shouldn't have to figure out what to do at 3am.
