Prometheus Alertmanager & Integrasi Notifikasi

groups:
  - name: server_infrastructure_alerts
    rules:
      # Peringatan 1: Server Down (Tidak dapat dihubungi)
      - alert: InstanceDown
        expr: up == 0
        for: 1m
        labels:
          severity: critical
        annotations:
          summary: "Server Instance {{ $labels.instance }} Mengalami Kerusakan (DOWN)"
          description: "Target {{ $labels.instance }} dari job {{ $labels.job }} tidak merespons selama lebih dari 1 menit."

      # Peringatan 2: CPU Server Terlalu Tinggi
      - alert: HighCpuUsage
        expr: 100 - (avg by(instance) (rate(node_cpu_seconds_total{mode="idle"}[2m])) * 100) > 85
        for: 2m
        labels:
          severity: warning
        annotations:
          summary: "Utilisasi CPU Tinggi pada {{ $labels.instance }}"
          description: "Penggunaan CPU melebihi 85% (Nilai saat ini: {{ $value | printf \"%.2f\" }}%)."

      # Peringatan 3: Kapasitas Harddisk Kritis (< 10%)
      - alert: DiskSpaceLow
        expr: (node_filesystem_free_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"}) * 100 < 10
        for: 5m
        labels:
          severity: critical
        annotations:
          summary: "Kapasitas Harddisk Menipis pada {{ $labels.instance }}"
          description: "Sisa ruang penyimpanan pada partisi root tersisa kurang dari 10%!"

global:
  resolve_timeout: 5m
  # Konfigurasi SMTP Email (Gmail / AWS SES)
  smtp_smarthost: 'smtp.gmail.com:587'
  smtp_from: 'alert-system@perusahaan.com'
  smtp_auth_username: 'emailkamu@gmail.com'
  smtp_auth_password: 'app-password-rahasia'
  smtp_require_tls: true

# Rute Penerusan Alert
route:
  group_by: ['alertname', 'cluster', 'service']
  group_wait: 30s        # Waktu tunggu pengumpulan batch alert pertama
  group_interval: 5m     # Jeda waktu pengiriman pembaruan grup
  repeat_interval: 4h    # Ulangi notifikasi setelah 4 jam jika masalah belum diperbaiki
  receiver: 'telegram-and-discord' # Default receiver

receivers:
  # Saluran 1: Telegram Bot & Discord Webhook
  - name: 'telegram-and-discord'
    telegram_configs:
      - bot_token: '123456789:ABCdefGhIJKlmNoPQRsTUVwxyZ' # Dapatkan dari @BotFather
        chat_id: -1001234567890                         # ID Grup Telegram Tim Ops
        parse_mode: 'HTML'
        send_resolved: true # Kirim pesan hijau saat server sudah pulih normal!
        message: |
          {{ if eq .Status "firing" }}🔥 <b>[ALERT - MASALAH TERDETEKSI]</b> 🔥{{ else }}✅ <b>[RESOLVED - MASALAH PULIH]</b> ✅{{ end }}
          <b>Peringatan:</b> {{ .CommonAnnotations.summary }}
          <b>Tingkat:</b> {{ .CommonLabels.severity }}
          <b>Deskripsi:</b> {{ .CommonAnnotations.description }}
          <b>Waktu:</b> {{ .StartsAt.Format "2006-01-02 15:04:05" }}

    discord_configs:
      - webhook_url: 'https://discord.com/api/webhooks/1234567890/token-rahasia'
        send_resolved: true
        title: '{{ if eq .Status "firing" }}:rotating_light: [ALERTING]{{ else }}:white_check_mark: [RESOLVED]{{ end }} {{ .CommonAnnotations.summary }}'
        message: '{{ .CommonAnnotations.description }}'

  # Saluran 2: Slack Webhook
  - name: 'slack-ops'
    slack_configs:
      - api_url: 'https://hooks.slack.com/services/T00/B00/XXXXX'
        channel: '#ops-alerts'
        send_resolved: true
        title: '[{{ .Status | toUpper }}] {{ .CommonAnnotations.summary }}'
        text: '{{ .CommonAnnotations.description }}'

  # Saluran 3: Email Notifikasi Tim Lead
  - name: 'email-admin'
    email_configs:
      - to: 'lead-devops@perusahaan.com'
        send_resolved: true