---
# Prometheus Alerting Rules для Memory Module
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: memory-module-alerts
  namespace: monitoring
  labels:
    app: memory-module
spec:
  groups:
    - name: memory_module
      interval: 30s
      rules:
        # NATS JetStream Alerts
        - alert: NATSOnlineBacklogHigh
          expr: nats_jetstream_stream_messages{stream="MM_ONLINE"} > 1000
          for: 5m
          labels:
            severity: critical
            component: nats
          annotations:
            summary: "MM_ONLINE backlog критично високий"
            description: "Backlog в MM_ONLINE stream: {{ $value }} messages. SLO порушено."
            
        - alert: NATSRedeliveriesSpike
          expr: rate(nats_jetstream_consumer_redeliveries_total[5m]) > 100
          for: 2m
          labels:
            severity: warning
            component: nats
          annotations:
            summary: "Спік redeliveries в NATS"
            description: "Redeliveries rate: {{ $value }}/min. Можливі проблеми з воркерами."
            
        - alert: NATSAckPendingHigh
          expr: nats_jetstream_consumer_ack_pending{stream="MM_ONLINE"} > 5000
          for: 5m
          labels:
            severity: warning
            component: nats
          annotations:
            summary: "Високий ack_pending в MM_ONLINE"
            description: "Ack pending: {{ $value }}. Воркери можуть бути перевантажені."
            
        - alert: NATSStreamStorageHigh
          expr: (nats_jetstream_stream_bytes / nats_jetstream_stream_max_bytes) > 0.8
          for: 10m
          labels:
            severity: warning
            component: nats
          annotations:
            summary: "Диск JetStream майже заповнений"
            description: "Використання: {{ $value | humanizePercentage }}"
            
        # Worker Alerts
        - alert: WorkerOffline
          expr: time() - worker_last_heartbeat_seconds > 120
          for: 2m
          labels:
            severity: critical
            component: worker
          annotations:
            summary: "Worker offline більше 2 хвилин"
            description: "Worker {{ $labels.node_id }} (Tier {{ $labels.tier }}) не відповідає."
            
        - alert: WorkerEmbedLatencyHigh
          expr: histogram_quantile(0.95, rate(worker_job_duration_seconds_bucket{type="embed"}[5m])) > 0.5
          for: 5m
          labels:
            severity: warning
            component: worker
          annotations:
            summary: "P95 latency для embed jobs > 500ms"
            description: "P95: {{ $value }}s (target: 300ms)"
            
        - alert: WorkerErrorRateHigh
          expr: rate(worker_errors_total[5m]) > 10
          for: 5m
          labels:
            severity: warning
            component: worker
          annotations:
            summary: "Високий error rate в воркерів"
            description: "Error rate: {{ $value }}/s"
            
        # Memory Service Alerts
        - alert: MemoryServiceDown
          expr: up{job="memory-service"} == 0
          for: 1m
          labels:
            severity: critical
            component: memory-service
          annotations:
            summary: "Memory Service недоступний"
            description: "Memory Service не відповідає на health checks."
            
        - alert: MemoryServiceLatencyHigh
          expr: histogram_quantile(0.95, rate(http_request_duration_seconds_bucket{job="memory-service"}[5m])) > 1.0
          for: 5m
          labels:
            severity: warning
            component: memory-service
          annotations:
            summary: "P95 latency Memory Service > 1s"
            description: "P95: {{ $value }}s"