UPS NUT and VMs

hspindel

Member
Aug 6, 2025
38
6
8
I have installed NUT as a client (of a Synology) into Proxmox VE. It looks like it works fine - upsc tells me good things, and systemctl status nut-client shows the daemon active and running.

Question: If power goes down, will PVE send shutdown commands to active VMs, or do I have to install nut-client in each VM?
 
will PVE send shutdown commands to active VMs, or do I have to install nut-client in each VM?
When issuing a shutdown command within the PVE host node, Proxmox will send a shutdown to all running VMs, but only if GA is configured (& running) within the VM will this be a regular shutdown command, otherwise an ACPI signal will be sent, so similar to pressing the Power button on a PC, while this will also "usually" work as a graceful shutdown, it may depend on the exact configuration of the VM. In all events, Proxmox will wait for a period defined as Shutdown timeout (default 180 seconds - can be changed), to let the VM gracefully shutdown. See here. If this Shutdown timeout period has passed & the VM is still running, Proxmox will force-stop this VM.
 
I've always installed GA in all my VMs
That's definitely the best approach - albeit not always possible in some edge cases (older or unconfigurable software).

I will only add for info purposes, a description of my UPS/NUT setup (at my home).

I have a NUT server (only) setup on the PVE host. (Usually I would avoid making such changes or similar to a PVE host node itself, however in my case with the UPS being attached by USB to the host, I took the decision to avoid using passthrough to an LXC or VM for my purposes, so inevitably made these fully documented & limited changes to the host itself).

On the PVE host I have a VM with Home Assistant which has Nut client configured (Nut integration). With this I can monitor the UPS/POWER status & have automations configured to take applicable/desired actions, the primary one being a Shutdown command sent (by SSH / API - see here) to the PVE host. This works flawlessly.

I chose this setup, primarily so as to make as little changes as possible, but also to have complete granular control over the UPS management, & still basically non Nut-dependent.

This works excellently for mu use-case, YMMV.