#!/bin/bash
# System Health Monitoring Script
# Generated by Axiom - Autonomous System Administrator

TIMESTAMP=$(date '+%Y-%m-%d %H:%M:%S')
REPORT_FILE="/opt/axiom/logs/health-$(date '+%Y-%m-%d').log"
DATA_DIR="/opt/axiom/data"

# Create log header
echo "=== System Health Report ===" >> "$REPORT_FILE"
echo "Timestamp: $TIMESTAMP" >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# CPU Information
echo "--- CPU Usage ---" >> "$REPORT_FILE"
CPU_USAGE=$(top -bn1 | grep "Cpu(s)" | sed "s/.*, *\([0-9.]*\)%* id.*/\1/" | awk '{print 100 - $1}')
echo "CPU Usage: ${CPU_USAGE}%" >> "$REPORT_FILE"
LOAD_AVG=$(uptime | awk -F'load average:' '{print $2}')
echo "Load Average:${LOAD_AVG}" >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# Memory Information
echo "--- Memory Usage ---" >> "$REPORT_FILE"
free -h >> "$REPORT_FILE"
MEM_PERCENT=$(free | grep Mem | awk '{printf("%.1f"), $3/$2 * 100.0}')
echo "Memory Usage: ${MEM_PERCENT}%" >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# Disk Information
echo "--- Disk Usage ---" >> "$REPORT_FILE"
df -h / /boot >> "$REPORT_FILE"
DISK_PERCENT=$(df / | tail -1 | awk '{print $5}' | sed 's/%//')
echo "" >> "$REPORT_FILE"

# Network Statistics
echo "--- Network Statistics ---" >> "$REPORT_FILE"
ss -s >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# Active Services
echo "--- Critical Services Status ---" >> "$REPORT_FILE"
for service in nginx sshd firewalld chronyd; do
    STATUS=$(systemctl is-active $service)
    echo "$service: $STATUS" >> "$REPORT_FILE"
done
echo "" >> "$REPORT_FILE"

# Failed Services
echo "--- Failed Services ---" >> "$REPORT_FILE"
FAILED=$(systemctl list-units --state=failed --no-pager --no-legend | wc -l)
if [ "$FAILED" -gt 0 ]; then
    systemctl list-units --state=failed --no-pager >> "$REPORT_FILE"
else
    echo "No failed services" >> "$REPORT_FILE"
fi
echo "" >> "$REPORT_FILE"

# Recent Authentication Attempts
echo "--- Recent Auth Attempts (Last 10) ---" >> "$REPORT_FILE"
journalctl _SYSTEMD_UNIT=sshd.service -n 10 --no-pager | grep -i "accepted\|failed" >> "$REPORT_FILE" 2>&1 || echo "No recent auth events" >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# Security: Check for security updates
echo "--- Security Updates ---" >> "$REPORT_FILE"
dnf check-update --security -q | tail -10 >> "$REPORT_FILE" 2>&1 || echo "System up to date" >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# Alert thresholds
ALERTS=""
if (( $(echo "$CPU_USAGE > 80" | bc -l) )); then
    ALERTS="${ALERTS}[ALERT] High CPU usage: ${CPU_USAGE}%\n"
fi
if (( $(echo "$MEM_PERCENT > 85" | bc -l) )); then
    ALERTS="${ALERTS}[ALERT] High memory usage: ${MEM_PERCENT}%\n"
fi
if [ "$DISK_PERCENT" -gt 85 ]; then
    ALERTS="${ALERTS}[ALERT] High disk usage: ${DISK_PERCENT}%\n"
fi
if [ "$FAILED" -gt 0 ]; then
    ALERTS="${ALERTS}[ALERT] Failed services detected: ${FAILED}\n"
fi

# Write alerts to separate file if any
if [ -n "$ALERTS" ]; then
    echo "=== SYSTEM ALERTS ===" >> "$REPORT_FILE"
    echo -e "$ALERTS" >> "$REPORT_FILE"
    echo -e "$ALERTS" > "$DATA_DIR/latest-alerts.txt"
else
    echo "=== No alerts - System healthy ===" >> "$REPORT_FILE"
    echo "No alerts" > "$DATA_DIR/latest-alerts.txt"
fi

echo "========================================" >> "$REPORT_FILE"
echo "" >> "$REPORT_FILE"

# Keep summary for quick reference
cat > "$DATA_DIR/latest-health.json" <<EOF
{
  "timestamp": "$TIMESTAMP",
  "cpu_usage": "$CPU_USAGE",
  "memory_usage": "$MEM_PERCENT",
  "disk_usage": "$DISK_PERCENT",
  "failed_services": "$FAILED",
  "alerts": $([ -n "$ALERTS" ] && echo "true" || echo "false")
}
EOF

echo "Health check completed at $TIMESTAMP"