Hi everyone,

I'm sure I'm not the only one virtualizing unRaid using Proxmox, and I'm trying to reduce the host machine power consumption as much as possible.

During this process, I’ve noticed strange host CPU usage when unRaid spins down the disks.

From my observations, unRaid seems to be very "nervous" by nature—it doesn’t sit still.

The images cpu_1, cpu_2, and affinity_spun_down show the host CPU fluctuations I’m referring to as "nervous." I don’t have Docker, VMs, or any plugins installed—it’s as bare-bones as possible at the moment.

Here’s the catch... Compare the host CPU usage when the disks are spun down, again images cpu_1, cpu_2, and affinity_spun_down, versus the images cpu_3, cpu_4, and affinity_spun_up, which show lower host CPU usage when the disks are spun up. Say what ?

Why is unRaid using more host CPU when the disks are spun down? It doesn’t make any sense to me. And why is unRaid so "nervous"? Stay still, dammit! lol

In all seriousness, this seems like a.. bug ?? ...and I’d like to ask the community for suggestions. I allocated 100GB of dedicated 1G HugePages to unRaid, and CPU affinity is set to 8-15 in Proxmox using "host" as the CPU setting. I passthrough the onboard 8 Port sata controller, an asm1166 6 port pcie card and a 1TB nvme that I will use for ZFS L2ARC, it's just not part of my zfs pool atm. I’m not using a Proxmox network bridge. My Proxmox server also functions as my router (OPNsense), so I’ve opted for an Intel X710-T2L to enable SR-IOV and use virtual network cards on all my VMs.

 

I tried messing with link_power_management_policy ( unRaid default is max_performance ) , just is case, and while changing it to med_power_with_dipm or min_power_with_parcial does lower the watt meter reading by about 5w (very nice) with my hardware, It does nothing for the cpu behavior. (This should be user selectable on the new "Power Mode" menu btw, imo!! ).

 

There is more to this host cpu usage for unRaid but lets start here. One thing at a time. 

 

Thank you !

tower-diagnostics-20250307-1227.zip

Edited March 7, 20251 yr by adolfotregosa

It's related to emhttpd. If I kill the process cpu usage goes down. Of course I can't use unraid correctly afterwards but why does that process use more cpu when the disks spindown?? 

My Proxmox server suffers from the same symptom. When the Unraid VM is turned off, the entire cluster's consumption stays around 80W. When Unraid is running, consumption fluctuates between 80W and 120W. I've already tried several things to reduce it but nothing works.

Edited November 23, 2025Nov 23 by Quejo