Server & Endpoint Monitoring
GridNMS can monitor Linux, macOS, and Windows machines the same way it monitors network devices: it connects over SSH with a dedicated, read-only account and reads CPU load, memory, disk usage, and uptime directly — no software to install on the machine being monitored.
This page covers setting up the read-only account on each operating system and importing the pack that reads it. Once both are done, monitoring the machine works like monitoring any other device: add it under Devices, put it in the matching class (Linux Server, macOS, or Windows Server), and enter the account’s username and password.
How it works
Section titled “How it works”GridNMS logs in with the read-only account on a schedule, runs a handful of built-in, harmless commands (reading memory/CPU stats, listing disk usage), and reads the results — the same way an administrator would type those commands by hand. Nothing is written back to the machine, and the account never needs administrator/root privileges for standard monitoring.
Linux hosts use the same built-in commands available on virtually every
distribution (/proc, df), so no special software is required beyond an
SSH server, which most Linux servers already run.
- Create a dedicated, unprivileged user for monitoring, for example:
Terminal window sudo useradd -m -s /bin/bash gridnms-rosudo passwd gridnms-ro - Make sure the account can log in over SSH (it can, by default, unless your
SSH server restricts login to specific groups — if it does, add
gridnms-roto the allowed group). - In GridNMS, add the device under the Linux Server class and enter
gridnms-roand its password under that class’s SSH settings (or override it per device).
No sudo access is required — everything GridNMS reads is available to any
logged-in user.
- Create the account. Open System Settings → Users & Groups, click
Add Account, and choose Standard (not Administrator). Give it a
name like
gridnms-roand a strong password. A Standard account is sufficient — GridNMS never needs admin rights. - Turn on Remote Login. Open System Settings → General → Sharing, and enable Remote Login. Under “Allow access for,” choose Only these users and add the account you just created — this keeps the SSH door closed to every other account on the Mac.
- Verify it works from another machine before wiring it into GridNMS:
If you’re prompted for the password and land at a shell prompt, you’re done. If the connection is refused, double-check Remote Login is on and the account is in the allowed list.
Terminal window ssh gridnms-ro@<the-mac's-address> - In GridNMS, add the device under the macOS class and enter the account’s username and password under that class’s SSH settings (or override it per device).
Windows (OpenSSH Server)
Section titled “Windows (OpenSSH Server)”Windows monitoring currently requires the built-in OpenSSH Server feature to be enabled. Most Windows Server installs since Windows Server 2019 include it, but it isn’t turned on by default.
- Enable OpenSSH Server, if it isn’t already:
- Settings app: Settings → Apps → Optional Features → search for “OpenSSH Server” → Install.
- Or PowerShell (as Administrator):
Terminal window Add-WindowsCapability -Online -Name OpenSSH.Server~~~~0.0.1.0Start-Service sshdSet-Service -Name sshd -StartupType Automatic
- Create a dedicated, low-privilege local account:
Leave it in the default Users group — do not add it to Administrators. GridNMS only needs a plain, non-admin login.
Terminal window net user gridnms-ro <a-strong-password> /add - Verify it works from another machine:
You should land at assh gridnms-ro@<the-server's-address>
C:\Users\gridnms-ro>prompt. - In GridNMS, add the device under the Windows Server class and enter the account’s username and password under that class’s SSH settings (or override it per device).
What gets monitored
Section titled “What gets monitored”Once connected, GridNMS reports:
- CPU — the share of processor time in use, as a percentage (all three platforms), plus the recent load average (Linux/macOS)
- Memory — total and free RAM, plus swap on Linux/macOS
- Disk usage — percent used per mount point (Linux/macOS) or drive letter (Windows)
- Network throughput — bytes in and out per network interface (Linux/macOS)
- Disk throughput — bytes read and written, per physical disk on Linux and for the machine as a whole on macOS
- Uptime (Linux/macOS)
- OS version — shown on the device’s overview
- Make and model — read from the machine’s own hardware information (Linux/macOS), so an agentlessly monitored server identifies itself on the device list the same way a switch or router does
These appear on the device’s detail page the same way any other monitored metric does, and can be charted, thresholded, and alerted on like any other data point.
On Linux and macOS, these are drawn as charts on the device’s System Stats tab and on Explore → Device Metrics:
- CPU — percent of processor time in use.
- Disk Usage — one line per mount point, showing the fullest few.
- Disk I/O — bytes read and written per second on one chart. On Linux that’s a read line and a write line for each disk, showing the busiest few; on macOS it’s a single read line and write line for the whole machine.
- Network — bits in and out per second for each network interface, showing the busiest few. To chart one interface on its own, pick it from the Interfaces section of Device Metrics — a Linux or Mac host monitored over SSH lists the interfaces it reports there by name, each with its own inbound and outbound Traffic chart.
- Load and Memory (with swap on its own axis), as before.
A Windows host is charted for CPU and Memory. Its disk usage is collected and can be charted, thresholded and alerted on from Explore → Device Metrics like any other reading, but it arrives without a chart of its own on the System Stats tab.
docs built 2026-11-12 · 945a2dec2