Observability and Health

Bridge Kernel continuously records runtime behavior for operations teams.

RuntimeMetricsSnapshot

Key fields:

• ingest_attempted
• ingest_accepted
• ingest_rejected_replay
• ingest_rejected_policy
• sync_published
• sync_received
• queue_peak

Event log format

JSONL records with:

• ts
• event
• detail

RuntimeHealthSnapshot

• kernel_id
• environment
• status
• degraded_reasons[]

Degradation signals

• queue_depth_high
• policy_denies_detected
• replay_rejections_detected

Recommended alerts

• Runtime status remains degraded for more than N minutes.
• Replay rejections exceed baseline for a source.
• Policy denials spike after policy promotion.
• Publish-receive sync imbalance persists.

Runbook hints

• Replay spike: inspect source clock drift and parser timestamps.
• Policy spike: diff current policy bundle vs previous active.
• Queue spike: inspect adapter throughput and storage latency.
• Sync lag: inspect inbound/outbound file movement and peer health.