metrics

Imported from zerokrab/bento-hancho (crates/hancho/src/metrics/AGENTS.md). Install upstream with npx skills add zerokrab/bento-hancho --skill metrics. Copyright stays with the author.

Metrics Implementation Notes

Separate metric groups for HTTP/API (metrics/api.rs), worker + job/taskdb (metrics/worker.rs), and request-level/request-queue (metrics/request_queue.rs), all wired through metrics/mod.rs::Metrics.

• DB is the source of truth for counts:
  • Worker online/offline gauges (worker_online_count, worker_offline_count) are derived from the workers table via HanchoSQLiteDB::get_worker_state_counts.
  • Per-worker job gauges (worker_running_jobs_count{worker_id}, worker_completed_jobs_count{worker_id}, worker_failed_jobs_count{worker_id}) are derived from worker_jobs via get_worker_job_state_counts.
  • Request metrics (request_state_count{state}) are derived from the requests table via get_request_state_counts.
• Where metrics are synced on each scrape:
  • api/rest_api.rs::get_metrics calls:
    • worker_registry.sync_metrics(&state.metrics) for worker + job gauges.
    • request_queue.sync_metrics(&state.metrics) for request state gauges.
• Where to add new metrics:
  • HTTP/API-level metrics → extend metrics/api.rs and plumb through Metrics.
  • Worker/taskdb-level metrics → extend metrics/worker.rs and expose helpers on WorkerMetrics.
  • Request-/RequestQueue-level metrics → extend metrics/request_queue.rs and update RequestQueue::sync_metrics.