Distributed Power Management (DPM)-style functionality in Proxmox VE?

[erikg]

Hello everyone,
I’m exploring ways to reduce power consumption in my Proxmox cluster, and I wanted to ask if there is already a feature (or recommended approach) similar to VMware’s Distributed Power Management (DPM).

Use case:

• I run a Proxmox cluster with 6 nodes.
• Most of the time, the cluster is under low load, and I don’t need all nodes running 24/7.
• Ideally, when the cluster has spare capacity, workloads could be consolidated onto fewer nodes, and the unused nodes could be powered down.
• If additional resources are needed, or if an HA event occurs (a node fails), Proxmox could use Wake-on-LAN or IPMI to bring another node online automatically.

My questions are:

1. Does Proxmox VE currently support anything like this natively?
2. If not, are there best practices or community tools/scripts that people are using to achieve it?
3. Would this be a candidate for a future feature—perhaps integrated into Proxmox HA or resource scheduling?

I know some people have used external scripts with Wake-on-LAN, and quorum devices (QDevice) to handle clusters where not all nodes are powered on. But I’m wondering if there’s a more “official” or supported direction here.
If this isn’t on the roadmap, I’d be very interested in contributing ideas or helping test, since energy-aware clustering seems increasingly relevant for homelabs and SMB deployments.
Thanks for your help!

 

[LnxBil]

Does Proxmox VE currently support anything like this natively?

Not at the moment.

Would this be a candidate for a future feature—perhaps integrated into Proxmox HA or resource scheduling?

Sure and a nice addition for big environments.

If this isn’t on the roadmap, I’d be very interested in contributing ideas or helping test, since energy-aware clustering seems increasingly relevant for homelabs and SMB deployments.

The smallest sane, energy efficient and proper cluster is a 3-node ceph cluster, which works fine for a lot of customers of various sizes, YMMV.
The "proper" here is high availability, infrastructure fault tolerence and trying to get rid of all SPOFs. I would never want to reduce this further.