Motivation

I need to maintain more than 20 physical servers at work. I connect to them via SSH and have to access each one individually to check CPU usage, disk usage, GPU usage, etc.

This process is time-consuming and doesn't scale well, so I built a tool to handle it.

Introducing ssh-monitor

ssh-monitor is a terminal tool that discovers SSH hosts from your config and shows real-time metrics in an interactive view.

Installation

Start the SSH agent and add your keys:

ssh-agent
ssh-add ~/.ssh/id_rsa

Install ssh-monitor:

curl -sSL https://raw.githubusercontent.com/tsugumi-sys/ssh-monitor/main/install.sh | bash

Run it:

ssh-monitor

Current Features

It auto-discovers hosts from your SSH config, uses a Ratatui-based UI for smooth navigation, and tracks CPU, memory, disk, and GPU metrics when available.

How It Works

Run ssh-monitor; it discovers hosts, shows a navigable list, and opens real-time charts for a selected host. CPU timelines, memory, disk, and GPU metrics update continuously.

Technical Implementation: Efficient SSH Connections

Minimizing SSH connections was the main challenge. Opening one session per metric would mean 120 connections per minute for 10 servers checked every five seconds. Instead, ssh-monitor batches metrics in a single session:

echo __BEGIN_cpu__
bash -c '
  if [[ "$(uname)" == "Darwin" ]]; then
    sysctl -a;
  else
    lscpu;
  fi
'
echo __END_cpu__
echo __BEGIN_mem__
bash -c 'uname -s && echo __MEM__ && (free -m || (echo __MAC__ && sysctl -n hw.memsize && vm_stat))'
echo __END_mem__
echo __BEGIN_disk__
df -Pm | tail -n +2
echo __END_disk__

Delimiter tags separate metrics so the tool can parse everything from one command run, cutting connections from 120 to about 30 per minute.