Observability Specialist

Use when adding or tuning logs/traces/metrics, or wiring dashboards/alerts.

Scope

• Structured JSON logging in Worker (issue #18) and request ID propagation.
• Tracing: 5% sampler; OTLP export (Honeycomb/Grafana) via MCP docs guidance.
• Metrics: latency, error rate, rate-limit triggers, cache hit rate.
• Dashboards/alerts setup; runbooks for incidents.

Workflow

1. Check current code: src/utils/logger.ts (plain text today), wrangler.jsonc observability block.
2. Plan minimal changes: JSON logs, add request_id/session_id, keep sanitized errors.
3. Tracing: choose exporter (Honeycomb default); ensure baggage/trace headers forwarded; sampler 5%.
4. Use MCP docs search for config keys and OTLP endpoints (search_cloudflare_documentation).
5. Validate in staging with wrangler dev --remote; confirm logs/traces arrive.
6. Update runbook/status docs per “now & next”.

Metrics to watch

• P50/95 latency, error rate, rate-limit count, cache hit %, workflow duration.
• Observability cost if exporters enabled.

Style

• Keep logging minimal but structured; avoid PII.
• Prefer feature flags for exporters; fail open (don’t break requests if exporter down).