#!/bin/bash
set -e

# Runaway Process Killer - Installation Script
# Protects against CPU and RAM runaway processes

echo "============================================"
echo "Runaway Process Killer - Installation"
echo "============================================"
echo ""

# Check root
if [ "$EUID" -ne 0 ]; then
    echo "Error: Please run as root (sudo ./install.sh)"
    exit 1
fi

# Detect script directory
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"

# Configuration
CPU_CYCLES="${CPU_CYCLES:-5}"        # 5 minutes default (5 cycles * 60s)
RAM_THRESHOLD="${RAM_THRESHOLD:-5}"  # 5% free memory threshold
SWAP_THRESHOLD="${SWAP_THRESHOLD:-5}" # 5% free swap threshold

echo "Configuration:"
echo "  CPU kill threshold: ${CPU_CYCLES} minutes of 95%+ CPU"
echo "  RAM kill threshold: <${RAM_THRESHOLD}% free memory"
echo ""

# Install packages (noninteractive to prevent hanging on prompts)
echo "[1/8] Installing packages..."
export DEBIAN_FRONTEND=noninteractive
dpkg --configure -a 2>/dev/null || true  # Fix any interrupted dpkg state
apt-get update -qq
apt-get install -y -qq earlyoom monit

# Configure earlyoom
echo "[2/8] Configuring earlyoom (RAM protection)..."
cat > /etc/default/earlyoom << EOF
# earlyoom configuration - managed by runaway-process-killer
EARLYOOM_ARGS="-m ${RAM_THRESHOLD} -s ${SWAP_THRESHOLD} --avoid '(^|/)(init|systemd|sshd)\$' -r 60"
EOF

# Configure monit daemon interval
echo "[3/8] Configuring monit daemon..."
if grep -q "^set daemon" /etc/monit/monitrc; then
    sed -i 's/^set daemon.*/set daemon 60/' /etc/monit/monitrc
else
    sed -i '1i set daemon 60' /etc/monit/monitrc
fi

# Enable monit HTTP interface (required for monit status)
if ! grep -q "^set httpd port 2812" /etc/monit/monitrc; then
    cat >> /etc/monit/monitrc << 'EOF'

# HTTP interface for monit status command
set httpd port 2812 and
    use address localhost
    allow localhost
    allow admin:monit
EOF
fi

# Create CPU killer config
echo "[4/8] Creating CPU monitoring rule..."
cat > /etc/monit/conf.d/cpu-killer << EOF
check system \$HOST
  if cpu usage > 95% for ${CPU_CYCLES} cycles then exec "/usr/local/bin/kill-top-cpu.sh"
EOF

# Create kill script
echo "[5/8] Installing CPU kill script..."
cat > /usr/local/bin/kill-top-cpu.sh << 'EOF'
#!/bin/bash
# Kill the process tree using the most CPU (excluding critical ones)

# Find the top CPU consumer (excluding protected and transient processes)
TOP_LINE=$(ps -eo pid,comm,%cpu --sort=-%cpu | grep -v -E '(PID|systemd|sshd|monit|earlyoom|bash|ps|awk|grep|head)' | head -1)
TARGET_PID=$(echo "$TOP_LINE" | awk '{print $1}')
COMM=$(echo "$TOP_LINE" | awk '{print $2}')

if [ -n "$TARGET_PID" ] && [ -n "$COMM" ]; then
    logger "monit cpu-killer: Killing all '$COMM' processes (detected high CPU on PID $TARGET_PID)"
    # Kill all processes with this command name
    pkill -9 -x "$COMM"
fi
EOF
chmod +x /usr/local/bin/kill-top-cpu.sh

# Create orphan Claude killer script
echo "[6/8] Installing orphan Claude killer script..."
cat > /usr/local/bin/kill-orphan-claude.sh << 'EOF'
#!/bin/bash
# Kill orphaned Claude processes (no TTY or parent is init)
# An orphaned Claude process is one where:
#   - TTY is "?" (no terminal attached), OR
#   - Parent PID is 1 (adopted by init)

LOG_TAG="orphan-claude-killer"

# Find all claude processes
CLAUDE_PIDS=$(pgrep -x claude 2>/dev/null)

if [ -z "$CLAUDE_PIDS" ]; then
    exit 0
fi

for PID in $CLAUDE_PIDS; do
    # Get TTY and PPID for this process
    PROC_INFO=$(ps -o tty=,ppid= -p "$PID" 2>/dev/null)

    if [ -z "$PROC_INFO" ]; then
        # Process already gone
        continue
    fi

    PROC_TTY=$(echo "$PROC_INFO" | awk '{print $1}')
    PARENT_PID=$(echo "$PROC_INFO" | awk '{print $2}')

    ORPHANED=false
    REASON=""

    # Check if TTY is "?" (no terminal)
    if [ "$PROC_TTY" = "?" ]; then
        ORPHANED=true
        REASON="no TTY attached"
    fi

    # Check if parent PID is 1 (adopted by init)
    if [ "$PARENT_PID" = "1" ]; then
        ORPHANED=true
        REASON="parent is init (PPID=1)"
    fi

    if [ "$ORPHANED" = true ]; then
        # Get process start time for logging
        START_TIME=$(ps -o lstart= -p "$PID" 2>/dev/null)
        CPU=$(ps -o %cpu= -p "$PID" 2>/dev/null)
        MEM=$(ps -o %mem= -p "$PID" 2>/dev/null)

        logger "$LOG_TAG: Killing orphaned claude process PID=$PID ($REASON) started='$START_TIME' cpu=$CPU% mem=$MEM%"

        # Kill the process tree (claude may have child processes)
        pkill -9 -P "$PID" 2>/dev/null
        kill -9 "$PID" 2>/dev/null
    fi
done
EOF
chmod +x /usr/local/bin/kill-orphan-claude.sh

# Create orphan Claude killer config
echo "[7/8] Creating orphan Claude monitoring rule..."
cat > /etc/monit/conf.d/orphan-claude-killer << 'EOF'
check program orphan-claude-killer with path "/usr/local/bin/kill-orphan-claude.sh"
  every 15 cycles
  if status != 0 then alert
EOF

# Enable and start services
echo "[8/8] Enabling and starting services..."
systemctl enable earlyoom monit
systemctl restart earlyoom monit

# Verify
echo ""
echo "============================================"
echo "Installation Complete!"
echo "============================================"
echo ""
echo "Status:"
systemctl is-active earlyoom && echo "  earlyoom: running" || echo "  earlyoom: FAILED"
systemctl is-active monit && echo "  monit: running" || echo "  monit: FAILED"
echo ""
echo "Configuration:"
echo "  CPU: Kill after ${CPU_CYCLES} minutes of 95%+ usage"
echo "  RAM: Kill when free memory < ${RAM_THRESHOLD}%"
echo "  Orphan Claude: Check every 15 minutes"
echo ""
echo "Commands:"
echo "  sudo monit status        # Check monit status"
echo "  journalctl -u earlyoom   # View earlyoom logs"
echo "  tail -f /var/log/monit.log  # View monit logs"
echo ""