Context

An e-commerce platform runs as a distributed system with multiple services. Operations engineers must detect and assess production anomalies quickly — without accessing source code or redeploying instrumentation.

Trigger

A production anomaly occurs (e.g., elevated error rate, latency spike, resource exhaustion, or unexpected traffic pattern).

Acceptance Criteria

• The anomaly is visible on operational dashboards within 2 minutes of onset
• All services emit structured logs (JSON) with at minimum: timestamp, severity, service name, trace ID, and error category
• All inter-service calls are captured as distributed traces with end-to-end latency breakdown per hop
• Key runtime metrics (request rate, error rate, p95 latency, CPU, memory) are collected at ≤ 30-second granularity
• Dashboards allow drill-down from system-wide overview to individual service metrics without custom queries
• Alerting rules trigger automated notifications within 90 seconds of threshold breach