production-monitoring-setup

Define critical metrics with thresholds across application, infrastructure, database, and dependency layers, then configure severity-tiered alerts and escalation procedures. This skill generates a hypercare_monitoring_setup.md covering dashboards, alert rules, and review schedules.

Track application performance (throughput, P95/P99 latency, error rates), infrastructure (CPU, memory, disk), database (query time, connection pool, replication lag), external dependencies, and business metrics like transaction success rates.

The documentation covers Datadog dashboard and alert rule syntax, Prometheus PromQL queries, and CloudWatch alarm definitions. Configurations include concrete examples like error rate alerts and latency thresholds.

Use four tiers: P0 for outages requiring 15-minute response and paging, P1 for significant degradation within 30 minutes, P2 for minor issues within 2 hours, and P3 for warning trends reviewed next business day.

You need the optimization report for performance baselines, the target architecture design for system components, the testing strategy for SLAs, plus team on-call structure and available monitoring platforms.