Windows 10 VM - Disk 100%

[archedraft]

Does anyone have any suggestions on how to improve the Disk performance of a Windows 10 VM? The VM is on a WD Red hard drive (WDC_WD30EFRX) and the VM is the only file accessing this hard drive. I took a screenshot of the Read and Write Speed which seems pretty slow. Windows 10 VM was created using unRAIDs default template (see screenshot).

[JonathanM]

Use a SSD instead of a spinner.

[archedraft]

Use a SSD instead of a spinner.

 

Indeed, my main VM is on a SSD and does not seem to have this issue. Unfortunately it doesn’t have enough space to share. Does there happen to be a known issue with spinners? I wasn’t expecting the spinner to be stellar but I was hoping for a little more...

[cuutip]

hmmm. I'm not sure that is all. I would add when I started my windows 10 / windows server VMs under UnRaid 6.4.1, the response was amazing. To add some more clarification, (this is all on spinners BTW, and they were 5400 rpms) I would start the VMs, they would take around 2 to maybe 5 minutes to fully load. I did not experience 100% disk usage issues during the load. To continue beating the horse. If I were to restart/reboot, poweroff/poweron a VM after a complete load, the VM would load in 10 - 15 seconds flat. And I noticed my Windows server loaded in 7 seconds. it was beautiful. I did not start to have issues with the Windows VMs riding the 100% disk usage until I upgraded to 6.8.3. Maybe this is a bug in the KVM, or mayber the implentation?

 

That is one of the reasons for my late upgrade. and quite possibly a reason for me to backport to 6.4.1. All my VMs Windows/Linux/BSD etc

responded much quicker under the KVM version included in 6.4.1.

 

I also notice there is a much greater resource use with the VMs, and more 100% processor spikes while monitoring VM usage with htop in the 6.8.3 distribution. in 6.4.1 the spikes are minimal and the resources do not seem to be taxed as much, and/or as often. 

[PeteAsking]

Most likely spinners are really bad for VM's because you are most likly using a journalised filesystem on top on another journalised filesystem. Its also so coincidental that you mention the WDC_WD30EFRX drive because that was one of the model numbers in the recent smr/cmr debacle. I did double check and that is a cmr so you are safe there at least.

[Jcloud]

I know the original post is old, but this could very well not be of any fault of Unraid or pass-through.  I've seen 100% disk usage on Window10 systems all the time, at my shop.  Typically you can click on Processes tab, and then sort by disk usage. Often times I've found Window services to be the culprit (indexing/searching, WMI, fault-reporting/diagnostics).  Other times it's a combination of dirty file system; windows files; registry; or multiple installed antivirus applications.

 

Easy first things to try:

1. Run from Admin-command prompt: 

chkdsk /f /x c:

Say yes to the prompt, then restart your VM/Windows. This will check/repair the file system.  If you suspect your drive has bad blocks use:

chkdsk /R /X C:

But this will take longer.

 

2.  Next from Admin-command prompt: 

sfc /scannow

This can be a hit or miss utility, in all meanings. If it finds and fixes errors. Reboot Windows and re-run command; if it comes back saying no errors it worked the first time.  If it says it found errors, again, it won't be able to repair what it finds. 

 

3.  If Search indexer service is going crazy, it could be the index is corrupt. To reset the search index (if it's on):

1.  Open "Control Panel" - change it to icon view.
2.  Click on "Indexing Options"
3. Click on "Advanced"
4. Click on the "Rebuild" button and click OK.

4. If the diagnostics and feedback service is spiking you can try:

1. Open "Settings" from Windows button.
2. Click on "Privacy"
3. Click on "Diagnostics & feedback"
4. Scroll towards bottom click on "Delete" button for "Delete diagnostic data" and set "Feedback frequency" to "Never."
5. Run Disk Cleanup as administrator -  Thumbnails, Recycle bin, and downloads are optional purge everything else.

 

I have more suggestions/steps but given this is an 1-year old post, and the my next batch of suggestions carry greater amounts of risk I'll hold off for now.