[SOLVED] UPS: Install NUT server in VM good idea or not?

[LooneyTunes]

Hi,

I am setting up an UPS and will use Network UPS Tools (NUT) to manage it. My goal is to have the VM's shut down in a specific sequence;

1. @ remaining power ~ 50%, shut down VMs x & y.
2. @ remaining power ~ 2 - 5%, shut down VM z and the Proxmox server.

I have read the Proxmox manual for VM start & shutdown capabilities, and as I have understood it, Proxmox receives a shut down command, and all VMs are then shut down, according to set Proxmox options for each machine, and finally the server itself.

I am following an excellent instruction from Techno Tim, but he is using a Raspberry Pi instead of a VM, to have monitoring of the UPS until power is depleted (or almost).

I was hoping it would be possible to do this using a Proxmox VM. So I guess the question here is, if there is any way to make this a bit more granular, so shutdown (and subsequently also the Proxmox server) of VMs could be triggered by remaining UPS power?

Thanks

 

[LnxBil]

I was hoping it would be possible to do this using a Proxmox VM. So I guess the question here is, if there is any way to make this a bit more granular, so shutdown (and subsequently also the Proxmox server) of VMs could be triggered by remaining UPS power?

Nothing directly, you have to write your commands yourself, but shutting down via pvesh is very simple and only needs SSH.

If you mean controlled shutdown, this works perfectly, but I double that stopping a few VMs will save you a lot of power, the baseline idle power demand is much higher, so that you need to shutdown nodes in order to improve your on-battery-time.

I am following an excellent instruction from Techno Tim, but he is using a Raspberry Pi instead of a VM, to have monitoring of the UPS until power is depleted (or almost).

Makes sense ... all monitoring and managing should be done from outside-of-the-cluster, also best to do that redundantly.

 

[LooneyTunes]

If you mean controlled shutdown, this works perfectly, but I double that stopping a few VMs will save you a lot of power, the baseline idle power demand is much higher, so that you need to shutdown nodes in order to improve your on-battery-time.

Well, shutting them down is not to preserve power, only to make sure they are shut down cleanly in an orderly fashion

Makes sense ... all monitoring and managing should be done from outside-of-the-cluster, also best to do that redundantly.

Agreed, tried to cut a corner here to avoid yet another device, but believe it's the correct route to take.

Thanks for your input!