Critical monitoring thresholds: 80% memory, 100ms latency, slow index rebuilds

Production monitoring requires alerts on memory usage above 80%, query latency exceeding 100ms (p99), and prolonged index rebuild operations to catch issues before cascading failures.

Set up Prometheus with Grafana dashboards. Configure alerts for: memory usage >80%, p99 query latency >100ms, and abnormally long index rebuild times. Focus on p99 latency, not p50.