prd-v08-monitoring-setup

# Monitoring Setup

Position in workflow: v0.8 Runbook Creation → **v0.8 Monitoring Setup** → v0.9 GTM Strategy

## Execution Mode

Default is **standard**. See [`.claude/rules/08-skill-execution-modes.md`](../../rules/08-skill-execution-modes.md) for selection logic.

| Mode | What this skill produces |
|------|--------------------------|
| **quick** | RED metrics on critical path only; 3–5 alerts linked to RUN-; single overview dashboard |
| **standard** | RED + USE coverage; SLOs for tier-1 services; full alert routing to RUN-; dashboards by audience |
| **deep** | Layered coverage (RED + USE + business + UX); multi-tier SLOs with error budgets; baseline calibration from staging; escalation routing |

## Consumes

This skill requires prior work from v0.8 Runbook Creation and earlier stages:

- **RUN-\* runbook entries** (from v0.8 Runbook Creation) — Incident response runbooks define alerting scenarios; critical alerts must link to RUN- procedures
- **DEP-\* deployment entries** (from v0.8 Release Planning) — DEP- rollback thresholds and post-deploy validation steps inform MON- alert conditions and SLO targets
- **API-\* endpoint contracts** (from v0.6 Technical Specification) — Define baseline latency, throughput, and error rates for application-layer metrics
- **KPI-\* metrics** (from v0.3 Outcome Definition and v0.9 Launch Metrics) — Business metrics (signups, conversions, retention) inform dashboard design and business layer monitoring
- **ARC-\* architecture decisions** (from v0.6 Architecture Design) — System structure determines which components to monitor (monolith has different metrics than distributed services)
- **TECH-\* technology stack** (from v0.5 Technical Stack Selection) — Technology choices (database, cloud provider, APM tools) determine available metrics and monitoring tools

This skill assumes DEP- and RUN- entries are complete with thresholds, rollback conditions, and incident procedures defined.

## Produces

This skill creates/updates:

- **MON-\* entries** (monitoring specifications, metric/alert/dashboard/SLO types) — Concrete monitoring rules with thresholds, alert conditions, dashboards, SLO definitions, linked to RUN- procedures
- **Alert routing configuration** — Mapping of MON- alerts to notification channels and teams; links alerts to RUN- incident procedures
- **Observability baseline** — Metrics gathered from staging/production, establishing normal operating ranges for alert thresholds

All MON- entries are **operational monitoring specifications**, not confidence-based. They are:
- **Measurable** (every metric has a source, unit, and aggregation method)
- **Actionable** (every alert has a RUN- procedure; no orphaned alerts)
- **Thresholded** (critical/warning severity with specific numeric conditions)
- **Dashboarded** (MON- dashboard entries provide visibility to operators and stakeholders)
- **SLO-backed** (SLO entries tie monitoring to product commitments)

Example MON- entries:
```markdown
MON-001: API Request Latency (p95)
Type: Metric
Layer: Application
Owner: Backend Team

Name: api.request.latency.p95
Description: 95th percentile response time for all API endpoints (from API-001–020)
Unit: ms
Source: Application APM (Datadog custom instrumentation)
Aggregation: p95 over 5-minute window
Retention: 90 days

Linked IDs: API-001 to API-020, DEP-004 (baseline from staging)

---

MON-002: High Latency Alert (Warning)
Type: Alert
Layer: Application
Owner: Backend Team

Metric: MON-001 (api.request.latency.p95)
Condition: >500ms (from DEP-002 baseline)
Window: 5 minutes
Severity: Warning
Runbook: RUN-001 (Performance Degradation Investigation)

Notification:
  - Channel: Slack #backend-alerts
  - Recipients: Backend on-call, team notified during business hours

Silencing: During scheduled maintenance windows (DEP-004 notifications)

Linked IDs: MON-001, RUN-001, DEP-002

---

MON-003: Critical Latency Alert
Type: Alert
Layer: Application
Owner: Backend Team

Metric: MON-001 (api.request.latency.p95)
Condition: >2000ms (SLA breach, from KPI-001 target)
Window: 2 minutes
Severity: Critical
Runbook: RUN-001 (Performance Degradation Investigation)

Notification:
  - Channel: PagerDuty (wake on-call)
  - Recipients: Backend on-call, Tech Lead, escalate if not acknowledged in 5 min

Silencing: None (critical alerts never silenced)

Linked IDs: MON-001, RUN-001, KPI-001

---

MON-004: API Availability SLO
Type: SLO
Layer: Application
Owner: Platform Team

Objective: API endpoints return non-5xx response
Target: 99.9% uptime (from DEP-002 / KPI-001)
Window: Rolling 30 days
Error Budget: 43.2 minutes/month

Alerting:
  - 50% error budget consumed → Warning to engineering (slow-burn alert)
  - 75% error budget consumed → Critical, freeze non-essential deploys
  - 100% error budget consumed → Post-incident review required (RUN-008 procedure)

Linked IDs: API-001–020, DEP-003 (rollback triggers), RUN-008 (incident review)

---

MON-005: System Health Dashboard
Type: Dashboard
Layer: Infrastructure + Application
Owner: Platform Team

Purpose: Quick health check for on-call engineers (run from RUN-002, RUN-001)
Audience: On-call engineers, engineering leadership, ops team
Panels:
  - API Request Rate (last 1h): Should be steady or increasing
  - API Latency (p50, p95, p99): Watch for p95/p99 creeping up
  - Error Rate by Endpoint: Any 5xx > 0 is concerning
  - Active Critical Alerts: Should be none
  - Database Connection Pool (from MON-006): Trending toward threshold
  - CPU/Memory by Service: Identify resource exhaustion
  - Deployment Status: Current version, time of last deploy
Refresh: 30 seconds

Linked IDs: MON-001, MON-002, MON-003, MON-006, DEP-001, RUN-001/002

---

MON-006: Database Connection Pool Utilization
Type: Metric
Layer: Infrastructure
Owner: Database Team

Name: db.connection_pool.utilized_percent
Description: Percentage of available connections in use (from DEP-001 pool size)
Unit: percentage
Source: Database monitoring (RDS Enhanced Monitor