sync: scripts/health_guard.sh

scripts/health_guard.sh

@@ -0,0 +1,44 @@
+#!/bin/bash
+# CROM-IA OOM/Swap Watchdog SRE Daemon
+
+LOG_FILE="../logs/health_guard.log"
+mkdir -p ../logs
+
+echo "[$(date)] SRE Health Guard Ativado." | tee -a "$LOG_FILE"
+
+while true; do
+    # Ler percentual de Swap usado (via free -m)
+    SWAP_INFO=$(free -m | grep Swap)
+    SWAP_TOTAL=$(echo "$SWAP_INFO" | awk '{print $2}')
+    SWAP_USED=$(echo "$SWAP_INFO" | awk '{print $3}')
+    
+    if [ "$SWAP_TOTAL" -gt 0 ]; then
+        SWAP_PCT=$(( 100 * SWAP_USED / SWAP_TOTAL ))
+    else
+        SWAP_PCT=0
+    fi
+
+    # Analisar RSS (Memoria Residente) do servidor Python
+    PYTHON_PID=$(pgrep -f "server.py" | head -n 1)
+    
+    if [ ! -z "$PYTHON_PID" ]; then
+        RSS_KB=$(ps -o rss= -p "$PYTHON_PID")
+        RSS_MB=$(( RSS_KB / 1024 ))
+        
+        # Guard: Se RSS Vazar acima de 2.5GB, abortar o motor pra salvar o Kernel
+        if [ "$RSS_MB" -gt 2500 ]; then
+            echo "[$(date)] [FATAL] Memory Leak Critico detectado no python (RSS: ${RSS_MB}MB)! Disparando SIGKILL. 🚀💥" | tee -a "$LOG_FILE"
+            kill -9 "$PYTHON_PID"
+        fi
+    fi
+
+    # Guard: Se SWAP passar de 96%, estamos prestes a travar = Abortar
+    if [ "$SWAP_PCT" -gt 96 ]; then
+        echo "[$(date)] [FATAL] Swap critical mass (96%). Destruindo montagens para resgatar Edge Device! 💥" | tee -a "$LOG_FILE"
+        fusermount -u ../mnt_crom 2>/dev/null || true
+        kill -9 "$PYTHON_PID" 2>/dev/null || true
+        exit 1
+    fi
+
+    sleep 5
+done