bp2008

Members
  • Posts

    7
  • Joined

  • Last visited

bp2008's Achievements

Noob

Noob (1/14)

8

Reputation

  1. TightVNC server, too, started making qemu act up.
  2. Continuing my monologue, I think I may have finally discovered the true cause of the elevated qemu-system-x86 CPU usage. It appears to have been Multiplicity 3's fault, somehow. The CPU% rises when I connect with Multiplicity KVM Mode, remains elevated after I disconnect, and falls back to normal only when I reboot the VM or when I kill all Multiplicity processes on it. I can only guess that Multiplicity may be failing to close a DirectX desktop duplication session or something. Anyway, I guess I'll learn to love VNC.
  3. I tried Hyper-V Server 2019 (it is a pain in the ass to configure by the way), and eventually got to spin up the same Win10 VM which I was running in unRAID. Overall CPU usage remained below 1% with the VM idle, and while that is good to see, power consumption actually wasn't any lower than when unRAID ran the same VM with elevated idle CPU usage. More importantly, CPU benchmarks in the VM scored only 50% to 66% of what I get in unRAID with the same numbers of cores assigned. Yikes! To add insult to injury, my preferred remote control software (Multiplicity 3) is unable to interact with the desktop when running under Hyper-V. I've decided that unRAID is back on the menu.
  4. So much for that. After behaving itself for over two weeks, idle CPU usage has elevated again and is now at 4-5% overall, with top showing ~48% CPU for qemu-system-x86 while the task manager in the win10 VM shows only 1% usage (that 1% is task manager!). I don't think I am willing to put up with kvm compatibility issues just so I can have unRAID's storage solution on my VM server.
  5. I decided to see if reducing the core count assigned to the VM would cause a proportional drop in idle CPU usage. I changed it from 14 logical CPUs to 8. OLD CONFIG: NEW CONFIG: I didn't expect to solve problem of too-high idle CPU usage just by reducing the pinned core count, but it has been running idle like this all day and top shows qemu-system-x86 using only 4-5% of one CPU core. That is quite reasonable for idle usage of a Win10 VM. Power consumption is down to 45 watts too (used to be 55 watts), according to the outlet the machine is connected to, so I know it isn't just a CPU usage measurement error.
  6. Hi everyone. I'm trying to make sense of the wacky CPU usage reporting I see in unRAID. The problem is, the dashboard is showing 6%-8% overall CPU load (most of the time, 6%) while my one and only VM (Win10 x64) is sitting idle at the desktop. If I open the VM in VNC and look at task manager's details tab, System Idle Process is at 99%. This is as idle as Windows 10 gets. top shows qemu-system-x86 using more than half a CPU core. This almost kinda sorta matches what the Dashboard is saying. But right there in the same top output, the line starting with %Cpu(s) shows 99.8% idle, 0.1% user space, 0.1% kernel space. These things are telling me different stories and I don't know what to make of it. I've also got a terminal there running htop, and its data is rather substantially different from Dashboard as well. Does anyone know why these numbers are so different? This is a fresh install of unRAID 5 days ago. Currently I am not using this system for file storage at all except the minimum required to get the hypervisor enabled, which meant setting up a single disk in the array. I have one virtual machine configured. It is Windows 10 x64 and has one drive, a SATA SSD which is mapped directly to this VM (it was originally installed on bare metal, and can boot outside of unRAID). The VM has 14 cores (or hyperthreads) assigned to it -- that is all of them except for the first (CPU 0 - HT 8). I previously had assigned all but the last core (CPU 7 - HT 15), and the problem was no different. So, I originally thought this was an issue with the VM configuration. I've searched for solutions and I've enabled the "hpet" timer in the VM's XML configuration, as is often recommended to fix high Win10 VM CPU usage. <clock offset='localtime'> <timer name='hpet' present='yes'/> <timer name='hypervclock' present='yes'/> </clock> It did not help in any way I've been able to discern. I've also tried a common variant of this XML config which specifies "rtc" and "pit" timers, to no significant effect. If I reboot this VM, the qemu-system-x86 usage tends to settle below 10% in the top output for a while, but it will eventually jump back up to ~60-75%. Here is the rest of the VM's configuration: https://i.imgur.com/KqWGXmj.png So, right now I can't decide if there is a serious efficiency problem with the current configuration, or if this is just something misleading like the CPU running at a low clock speed due to low demand, and consequently causing artificially high usage %.