# =============================================================================
# Prometheus Remote Write → SigNoz - Phase O-3
# =============================================================================
# 建立者: Claude Code (首席架構師)
# 日期: 2026-04-02 (台北時間)
#
# ❌ 此方案已廢棄 (2026-04-02 實際部署時發現)
# 原因: SigNoz OTEL Collector 不支援 Prometheus remote_write 格式 (Protobuf)
#       端點 /api/v1/write 回傳 404 Not Found
#
# ✅ 改用方案: SigNoz OTEL Collector Prometheus Receiver 直接 scrape
#   設定檔: ops/signoz/otel-collector-config-phase-o.yaml
#   實際部署: .188:/home/ollama/signoz/deploy/docker/otel-collector-config.yaml
#   新增 jobs: node-from-signoz (node-exporter) + kube-state-from-signoz
#
# =============================================================================

# ===== 新增至 prometheus.yml 最外層 (與 scrape_configs 同級) =====

remote_write:
  - url: http://localhost:24318/api/v1/write
    # 只在同一台機器上 (188 → 188)，無需跨網路
    remote_timeout: 30s

    # 佇列配置: 適合小型叢集
    queue_config:
      capacity: 2500
      max_shards: 5
      min_shards: 1
      max_samples_per_send: 500
      batch_send_deadline: 5s

    # ===== 白名單過濾 (只轉關鍵指標) =====
    # 預估 ~50 個 metric series，90 天約 2-5 GB
    write_relabel_configs:
      # 1. 節點資源指標
      - source_labels: [__name__]
        regex: "node_cpu_seconds_total|node_memory_MemAvailable_bytes|node_memory_MemTotal_bytes|node_filesystem_avail_bytes|node_filesystem_size_bytes|node_load1|node_load5|node_load15|node_network_receive_bytes_total|node_network_transmit_bytes_total"
        action: keep

      # 2. 容器資源指標
      - source_labels: [__name__]
        regex: "container_cpu_usage_seconds_total|container_memory_working_set_bytes|container_memory_rss|container_network_receive_bytes_total|container_network_transmit_bytes_total"
        action: keep

      # 3. K8s 狀態指標 (Pod 重啟是 RCA 關鍵！)
      - source_labels: [__name__]
        regex: "kube_pod_container_status_restarts_total|kube_pod_status_phase|kube_deployment_status_replicas_available|kube_deployment_status_replicas_unavailable|kube_node_status_condition"
        action: keep

      # 4. API 效能指標
      - source_labels: [__name__]
        regex: "http_request_duration_seconds.*|http_requests_total"
        action: keep

      # 5. AWOOOI 自訂業務指標
      - source_labels: [__name__]
        regex: "awoooi_.*"
        action: keep

      # 6. 資料庫指標
      - source_labels: [__name__]
        regex: "pg_stat_activity_count|pg_stat_database_conflicts|pg_stat_database_deadlocks|pg_up|redis_connected_clients|redis_memory_used_bytes|redis_keyspace_hits_total|redis_keyspace_misses_total|redis_up"
        action: keep

      # 7. Probe 健康狀態
      - source_labels: [__name__]
        regex: "probe_success|probe_duration_seconds|probe_ssl_earliest_cert_expiry"
        action: keep

      # 8. UP 指標 (所有 target 存活狀態)
      - source_labels: [__name__]
        regex: "up"
        action: keep

# =============================================================================
# ⚠️ 注意事項
# =============================================================================
#
# 1. write_relabel_configs 是 OR 邏輯：多個 keep 規則中，
#    任一匹配即保留。這是 Prometheus 內建行為。
#
# 2. 如果 SigNoz OTLP/HTTP 端點不支援 Prometheus remote_write 格式，
#    需要改用 victoria-metrics 或 cortex 作為中繼。
#    替代方案: 在 SigNoz 的 OTEL Collector 中啟用 prometheusremotewrite receiver。
#
# 3. 增加/移除指標後，用以下指令驗證:
#    curl -s "http://192.168.0.188:9090/api/v1/query?query=up" | jq '.data.result | length'
#