🚀 Auto-deploy: BotVPS atualizado em 21/04/2026 20:42:48

2026-04-21 20:42:48 +00:00
parent 0ab0fb826c
commit c538fdb9c2
4 changed files with 186 additions and 0 deletions
--- a/watchdog.py
+++ b/watchdog.py
@@ -0,0 +1,96 @@
+import os
+import psutil
+import time
+import json
+import httpx
+import subprocess
+from dotenv import load_dotenv
+
+load_dotenv()
+
+TOKEN = os.getenv("TELEGRAM_BOT_TOKEN")
+CHAT_ID = os.getenv("TELEGRAM_CHAT_ID")
+
+# Configurações do Watchdog
+CPU_THRESHOLD = 90.0
+CPU_STREAK_LIMIT = 6  # 6 * 10s = 60s
+CHECK_INTERVAL = 10  # segundos
+
+class Watchdog:
+    def __init__(self):
+        self.cpu_streak = 0
+        self.last_alert_time = 0
+        self.alert_cooldown = 300 # 5 minutos entre alertas do mesmo tipo
+
+    async def send_telegram_message(self, message: str):
+        if not TOKEN or not CHAT_ID:
+            print("[WATCHDOG] Erro: TOKEN ou CHAT_ID não configurados.")
+            return
+        
+        url = f"https://api.telegram.org/bot{TOKEN}/sendMessage"
+        payload = {
+            "chat_id": CHAT_ID,
+            "text": f"🚨 **[WATCHDOG VPS]**\n\n{message}",
+            "parse_mode": "Markdown"
+        }
+        
+        try:
+            async with httpx.AsyncClient() as client:
+                await client.post(url, json=payload)
+        except Exception as e:
+            print(f"[WATCHDOG] Erro ao enviar Telegram: {e}")
+
+    def get_pm2_status(self):
+        try:
+            result = subprocess.run(["pm2", "jlist"], capture_output=True, text=True)
+            if result.returncode == 0:
+                data = json.loads(result.stdout)
+                issues = []
+                for proc in data:
+                    if proc['pm2_env']['status'] != 'online':
+                        issues.append(f"🔴 App '{proc['name']}' está {proc['pm2_env']['status']}!")
+                return issues
+        except Exception as e:
+            print(f"[WATCHDOG] Erro PM2: {e}")
+        return []
+
+    async def run(self):
+        print("[WATCHDOG] Iniciado. Vigilância ativa...")
+        
+        while True:
+            try:
+                # 1. Monitoramento de CPU
+                cpu_usage = psutil.cpu_percent(interval=1)
+                if cpu_usage > CPU_THRESHOLD:
+                    self.cpu_streak += 1
+                else:
+                    self.cpu_streak = 0
+                
+                if self.cpu_streak >= CPU_STREAK_LIMIT:
+                    if time.time() - self.last_alert_time > self.alert_cooldown:
+                        await self.send_telegram_message(
+                            f"CPU em nível crítico: {cpu_usage}% por mais de 1 minuto!"
+                        )
+                        self.last_alert_time = time.time()
+                
+                # 2. Monitoramento de PM2
+                pm2_issues = self.get_pm2_status()
+                if pm2_issues:
+                    await self.send_telegram_message("\n".join(pm2_issues))
+                
+                # 3. Monitoramento de Espaço em Disco
+                disk = psutil.disk_usage('/')
+                if disk.percent > 95:
+                    if time.time() - self.last_alert_time > self.alert_cooldown:
+                        await self.send_telegram_message(f"Espaço em disco crítico: {disk.percent}% ocupado!")
+                        self.last_alert_time = time.time()
+
+                await asyncio.sleep(CHECK_INTERVAL)
+            except Exception as e:
+                print(f"[WATCHDOG] Erro no loop: {e}")
+                await asyncio.sleep(CHECK_INTERVAL)
+
+if __name__ == "__main__":
+    import asyncio
+    dog = Watchdog()
+    asyncio.run(dog.run())