freshStartScript/freshStart.sh

#!/bin/bash
set -e

echo "Starting Proxmox fresh installation script..."

# Cleanup function for failed installations
cleanup() {
    echo "Cleaning up on error..."
    systemctl stop node_exporter 2>/dev/null || true
    systemctl stop hwmon.timer 2>/dev/null || true
    systemctl disable node_exporter 2>/dev/null || true
    systemctl disable hwmon.timer 2>/dev/null || true
    rm -f /etc/systemd/system/node_exporter.service
    rm -f /etc/systemd/system/hwmon.service
    rm -f /etc/systemd/system/hwmon.timer
    rm -f /usr/local/bin/node_exporter
    rm -rf node_exporter-*.linux-amd64.tar.gz node_exporter-*.linux-amd64
    userdel node_exporter 2>/dev/null || true
    systemctl daemon-reload
    echo "Cleanup completed."
}

# Set trap for cleanup on error
trap cleanup ERR

# Install dependencies
echo "Installing required packages..."
apt-get update
apt-get install -y python3-pip smartmontools iperf3 python3-psutil python3-requests lm-sensors fastfetch rsync fio

# Install Node Exporter
echo "Installing Prometheus Node Exporter..."
NODE_EXPORTER_VERSION="1.8.2"

# Download Node Exporter with error handling
echo "Downloading Node Exporter..."
if ! wget --timeout=30 --tries=3 "https://github.com/prometheus/node_exporter/releases/download/v${NODE_EXPORTER_VERSION}/node_exporter-${NODE_EXPORTER_VERSION}.linux-amd64.tar.gz"; then
    echo "ERROR: Failed to download Node Exporter"
    exit 1
fi

# Extract with error checking
if ! tar xvfz node_exporter-*.linux-amd64.tar.gz; then
    echo "ERROR: Failed to extract Node Exporter archive"
    exit 1
fi

# Check if user already exists
if ! id "node_exporter" &>/dev/null; then
    echo "Creating node_exporter user..."
    useradd -rs /bin/false node_exporter
else
    echo "node_exporter user already exists, skipping creation..."
fi

# Move binary to proper location
if ! mv node_exporter-${NODE_EXPORTER_VERSION}.linux-amd64/node_exporter /usr/local/bin/; then
    echo "ERROR: Failed to move Node Exporter binary"
    exit 1
fi

# Cleanup downloaded files
rm -rf node_exporter-*.linux-amd64.tar.gz node_exporter-*.linux-amd64

# Set proper permissions
chown node_exporter:node_exporter /usr/local/bin/node_exporter
chmod +x /usr/local/bin/node_exporter

# Create node_exporter service file
cat > /etc/systemd/system/node_exporter.service << 'EOL'
[Unit]
Description=Node Exporter
After=network.target

[Service]
User=node_exporter
Group=node_exporter
Type=simple
ExecStart=/usr/local/bin/node_exporter

[Install]
WantedBy=multi-user.target
EOL

# Enable and start node_exporter
systemctl daemon-reload
systemctl enable node_exporter
systemctl start node_exporter

# Check if Node Exporter started successfully
if ! systemctl is-active --quiet node_exporter; then
    echo "ERROR: Node Exporter failed to start"
    systemctl status node_exporter --no-pager || true
    exit 1
fi

# Install hwmon daemon
echo "Installing hwmon daemon..."

# Download hwmon service files with error handling
echo "Downloading hwmon service files..."
if ! curl --max-time 30 --retry 3 -o /etc/systemd/system/hwmon.service http://10.10.10.63:3000/LotusGuild/hwmonDaemon/raw/branch/main/hwmon.service; then
    echo "ERROR: Failed to download hwmon.service"
    exit 1
fi

if ! curl --max-time 30 --retry 3 -o /etc/systemd/system/hwmon.timer http://10.10.10.63:3000/LotusGuild/hwmonDaemon/raw/branch/main/hwmon.timer; then
    echo "ERROR: Failed to download hwmon.timer"
    exit 1
fi

# Verify downloaded files exist and are not empty
if [[ ! -s /etc/systemd/system/hwmon.service ]]; then
    echo "ERROR: hwmon.service file is empty or missing"
    exit 1
fi

if [[ ! -s /etc/systemd/system/hwmon.timer ]]; then
    echo "ERROR: hwmon.timer file is empty or missing"
    exit 1
fi

# Create log directory for hwmon
mkdir -p /var/log/hwmonDaemon

# Start the hwmon daemon
systemctl daemon-reload
systemctl enable hwmon.timer
systemctl start hwmon.timer

# Check if hwmon timer started successfully
if ! systemctl is-active --quiet hwmon.timer; then
    echo "ERROR: hwmon timer failed to start"
    systemctl status hwmon.timer --no-pager || true
    exit 1
fi

# Final verification
echo "Verifying installation..."
echo "Node Exporter status: $(systemctl is-active node_exporter)"
echo "hwmon timer status: $(systemctl is-active hwmon.timer)"
echo "Node Exporter port check:"
if ss -tlnp | grep :9100; then
    echo "✓ Node Exporter is listening on port 9100"
else
    echo "WARNING: Node Exporter not listening on port 9100"
fi

# Test hwmon with error handling
echo "Testing hwmon dry-run..."
if ! /usr/bin/env python3 -c "import urllib.request; exec(urllib.request.urlopen('http://10.10.10.63:3000/LotusGuild/hwmonDaemon/raw/branch/main/hwmonDaemon.py').read().decode('utf-8'))" --dry-run; then
    echo "WARNING: hwmon dry-run test failed, but services are installed"
fi

# Disable cleanup trap on successful completion
trap - ERR

echo "✓ Installation complete! All services are running correctly."
echo ""
echo "Services installed:"
echo "  - Node Exporter: http://$(hostname -I | awk '{print $1}'):9100/metrics"
echo "  - hwmon daemon: Monitoring system health every 15 minutes"
echo ""
echo "Log locations:"
echo "  - Node Exporter: journalctl -u node_exporter"
echo "  - hwmon: journalctl -u hwmon.service"
echo "  - hwmon logs: /var/log/hwmonDaemon/"