Observability Coverage #38
Description
Why
Make issues diagnosable with minimal reproduction.
Definition of Done
- Structured logs exist for request spans, job spans, fetch, extract, and index actions.
- Metrics include request counts, error rates, durations, worker queue depth, job attempts, fetch sizes, and index size.
- Sampling and log levels are configurable per environment.
Tasks
- Add span fields and events to critical paths.
- Emit counters, gauges, and timers for the listed signals.
- Provide a default configuration file with sensible levels.
- Extend the runbook with troubleshooting queries.