15.255 14.9.15 Catálogo de métricas
| ID | Métrica | Tipo | Definición | Labels |
|---|---|---|---|---|
| MET-01 | capture_frames_total |
counter | Frames recibidos por stream y disposition. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-02 | capture_frame_gap_total |
counter | Saltos de sequence/timestamp detectados. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-03 | capture_fps_effective |
gauge | FPS observado por ventana. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-04 | capture_timestamp_skew_seconds |
histogram | Residual temporal entre streams correlacionados. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-05 | capture_write_latency_seconds |
histogram | Latencia de commit de medios. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-06 | capture_device_reconnect_total |
counter | Reconexiones de cámara. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-07 | depth_valid_fraction |
gauge | Fracción de depth válido por bundle. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-08 | queue_jobs |
gauge | Jobs por stage/state/priority con labels acotados. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-09 | queue_oldest_ready_age_seconds |
gauge | Edad del job ready más antiguo. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-10 | job_duration_seconds |
histogram | Duración por etapa y outcome. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-11 | job_retry_total |
counter | Reintentos por clase de error. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-12 | job_lease_expired_total |
counter | Leases expirados. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-13 | job_checkpoint_total |
counter | Checkpoints persistidos. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-14 | inference_invocations_total |
counter | Invocaciones por model family/version digest prefix. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-15 | inference_latency_seconds |
histogram | Latencia end-to-end y compute por modelo. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-16 | inference_batch_size |
histogram | Batch efectivo. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-17 | inference_oom_total |
counter | Fallos de memoria por worker/profile. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-18 | inference_fallback_total |
counter | Cambios a runtime/model fallback. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-19 | pose_valid_joint_fraction |
gauge | Cobertura de joints válidos. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-20 | scene_geometry_valid_fraction |
gauge | Cobertura geométrica válida. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-21 | contact_unknown_fraction |
gauge | Fracción de contacto desconocido. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-22 | phase_unknown_fraction |
gauge | Fracción temporal sin fase publicable. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-23 | classification_abstention_ratio |
gauge | Proporción de episodios con abstención. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-24 | quality_gate_total |
counter | Decisiones por gate/outcome. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-25 | result_publish_latency_seconds |
histogram | Capture committed a result published. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-26 | storage_bytes |
gauge | Bytes por data zone. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-27 | storage_free_bytes |
gauge | Espacio disponible. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-28 | storage_time_to_full_seconds |
gauge | Proyección robusta de capacidad. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-29 | storage_io_latency_seconds |
histogram | Latencia de IO por operation class. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-30 | artifact_digest_failure_total |
counter | Verificaciones de digest fallidas. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-31 | metadata_transaction_failure_total |
counter | Transacciones fallidas. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-32 | sync_pending_bytes |
gauge | Bytes pendientes de sincronización. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-33 | sync_object_failure_total |
counter | Fallos por clase de sync. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-34 | sync_last_success_age_seconds |
gauge | Edad desde sync exitoso. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-35 | telemetry_dropped_total |
counter | Señales descartadas por tipo/razón. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-36 | process_restart_total |
counter | Reinicios por servicio/causa. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-37 | watchdog_missed_heartbeat_total |
counter | Heartbeats perdidos. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-38 | jetson_ram_used_bytes |
gauge | Memoria usada derivada de OS/tegrastats. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-39 | jetson_swap_used_bytes |
gauge | Swap usada. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-40 | jetson_gpu_utilization_ratio |
gauge | Utilización GPU. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-41 | jetson_temperature_celsius |
gauge | Temperatura por sensor estable. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-42 | jetson_power_watts |
gauge | Potencia disponible/estimada por rails soportados. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-43 | jetson_thermal_throttle_total |
counter | Eventos de throttling. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-44 | api_request_duration_seconds |
histogram | Latencia de API por route template/status class. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-45 | api_requests_in_flight |
gauge | Requests concurrentes. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-46 | ui_review_revision_total |
counter | Revisiones humanas creadas. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-47 | model_verification_failure_total |
counter | Firmas/digests/compatibilidad inválidos. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
| MET-48 | deployment_rollback_total |
counter | Rollbacks por razón. | LowCardinalityOnly; never athleteId, attemptId, file path or full model digest as label |
Prometheus recomienda integrar instrumentación en cada subsistema y controlar la cardinalidad de labels. Por eso attemptId, athleteId, rutas completas y hashes completos quedan fuera del espacio de labels [P49-R20].