Proxmox/misc/monitor-all.sh

#!/usr/bin/env bash

# Copyright (c) 2021-2023 tteck
# Author: tteck (tteckster)
# License: MIT
# https://github.com/tteck/Proxmox/raw/main/LICENSE

clear
cat <<"EOF"
    __  ___            _ __                ___    ____
   /  |/  /___  ____  (_) /_____  _____   /   |  / / /
  / /|_/ / __ \/ __ \/ / __/ __ \/ ___/  / /| | / / /
 / /  / / /_/ / / / / / /_/ /_/ / /     / ___ |/ / /
/_/  /_/\____/_/ /_/_/\__/\____/_/     /_/  |_/_/_/

EOF

add() {
while true; do
  read -p "This script will add Monitor All to Proxmox VE. Proceed(y/n)?" yn
  case $yn in
  [Yy]*) break ;;
  [Nn]*) exit ;;
  *) echo "Please answer yes or no." ;;
  esac
done

echo '#!/usr/bin/env bash
# Read excluded instances from command line arguments
excluded_instances=("$@")
echo "Excluded instances: ${excluded_instances[@]}"

while true; do

  for instance in $(pct list | awk '\''{if(NR>1) print $1}'\''; qm list | awk '\''{if(NR>1) print $1}'\''); do
    # Skip excluded instances
    if [[ " ${excluded_instances[@]} " =~ " ${instance} " ]]; then
      continue
    fi

    # Determine the type of the instance (container or virtual machine)
    if pct status $instance >/dev/null 2>&1; then
      # It is a container
      config_cmd="pct config"
      IP=$(pct exec $instance ip a s dev eth0 | awk '\''/inet / {print $2}'\'' | cut -d/ -f1)
    else
      # It is a virtual machine
      config_cmd="qm config"
      IP=$(qm guest cmd $instance network-get-interfaces | egrep -o "([0-9]{1,3}\.){3}[0-9]{1,3}" | grep -E "192\.|10\.")
    fi

    # Skip instances based on templates
    template=$($config_cmd $instance | grep template | grep -q "template:" && echo "true" || echo "false")
    if [ "$template" == "true" ]; then
      echo "Skipping $instance because it is a template"
      continue
    fi

    # Ping the instance
    if ! ping -c 1 $IP >/dev/null 2>&1; then
      # If the instance can not be pinged, stop and start it
      if pct status $instance >/dev/null 2>&1; then
        # It is a container
        echo "$(date): CT $instance is not responding, restarting..."
        pct stop $instance >/dev/null 2>&1
        sleep 5
        pct start $instance >/dev/null 2>&1
      else
        # It is a virtual machine
        if qm status $instance | grep -q "status: running"; then
          echo "$(date): VM $instance is not responding, resetting..."
          qm reset $instance >/dev/null 2>&1
        else
          qm start $instance >/dev/null 2>&1
          echo "$(date): VM $instance is not running, starting..."
        fi
      fi
    fi
  done

  # Wait for 5 minutes. (Edit to your needs)
  echo "$(date): Pausing for 5 minutes..."
  sleep 300
done >> /var/log/ping-instances.log 2>&1' >/usr/local/bin/ping-instances.sh

# Change file permissions to executable
chmod +x /usr/local/bin/ping-instances.sh

# Create ping-instances.service
echo '[Unit]
Description=Ping instances every 5 minutes and restarts if necessary

[Service]
Type=simple
# Include the CT/VM ID at the end of the line where ExecStart=/usr/local/bin/ping-instances.sh is specified,
# to indicate which CT/VN should be excluded. Example: ExecStart=/usr/local/bin/ping-instances.sh 100 102
ExecStart=/usr/local/bin/ping-instances.sh
Restart=always
StandardOutput=file:/var/log/ping-instances.log
StandardError=file:/var/log/ping-instances.log

[Install]
WantedBy=multi-user.target' >/etc/systemd/system/ping-instances.service

# Reload daemon, enable and start ping-instances.service
systemctl daemon-reload
systemctl enable -q --now ping-instances.service
clear
echo -e "\n To view Monitor All logs: cat /var/log/ping-instances.log"
}

remove() {
  systemctl stop ping-instances.service
  systemctl disable ping-instances.service &>/dev/null
  rm /etc/systemd/system/ping-instances.service
  rm /usr/local/bin/ping-instances.sh
  rm /var/log/ping-instances.log
  echo "Removed Monitor All from Proxmox VE"
}

if [ "$1" == "add" ]; then
    add
elif [ "$1" == "remove" ]; then
    remove
else
    echo "Usage: $0 [ -s add | -s remove ]"
    exit 1
fi