Monitoring & Observability Expert

Production Monitoring Stack

Focus: Metrics, Logs, Traces (Three Pillars of Observability)
Stack: Prometheus, Grafana, Loki, OpenTelemetry, Jaeger

Overview

Complete observability for production systems.

Three Pillars

1. Metrics: Time-series data (Prometheus)
2. Logs: Event records (Loki, ELK)
3. Traces: Distributed request tracking (Jaeger, Tempo)

Quick Start

1. Metrics (Prometheus)

Counter, Gauge, Histogram for application metrics.

Key Metrics:

• Request rate (requests/sec)
• Error rate (5xx/total)
• Latency (p50, p95, p99)

See: examples.md

2. Logging (Structured)

JSON-formatted logs for easy parsing.

Fields: timestamp, level, message, context

See: examples.md

3. Tracing (OpenTelemetry)

Track requests across microservices.

Concepts: Span, Trace ID, Parent-Child relationships

See: examples.md

4. Alerting

Automated alerts on threshold breaches.

Examples: Error rate >5%, CPU >80%

See: examples.md

Monitoring Methodologies

RED Method (for Services)

• Rate: Requests per second
• Errors: Error percentage
• Duration: Latency distribution

USE Method (for Resources)

• Utilization: % busy
• Saturation: Queue depth
• Errors: Error count

Best Practices

1. Cardinality: Avoid high-cardinality labels (e.g., user_id)
2. Retention: 15 days (metrics), 7 days (logs)
3. Sampling: 1% trace sampling for high-traffic services
4. Dashboards: One dashboard per service

Validation Checklist

• Metrics: Prometheus scraping configured?
• Logs: Structured (JSON) format?
• Traces: OpenTelemetry instrumented?
• Alerts: Critical alerts defined?
• Dashboards: Grafana dashboards created?

Related Skills

• moai-essentials-perf: Performance profiling
• moai-devops-docker: Container monitoring
• moai-cloud-aws-advanced: CloudWatch

Additional Resources

• examples.md: Implementation code
• reference.md: Prometheus query language (PromQL)

Last Updated: 2025-11-20