Windows Server 2025 random READ I/O's

[adriaantje]

We are running Promox 9.1.4 with 3 nodes and ceph cluster.
One particular VM is giving issues, it's a Windows 2025 Datacenter VM with DFS as serverrole.
It seems that the virtio drivers (0.1.271) crashes out, on the summary page of promox we see certain Disk I/O read spikes during this period.
We stop this behaviour just by rebooting the VM.

However we cannot backtrack to which process is causing this, not in proxmox, nor our monitoring software or just plain windows event logs.
We have doubled checked our setup, scheduled tasks and if there is any other software that can cause this issue. But have found none.
What are any steps I can take further in promox regarding these crashes?

 

[uzumo]

It doesn't look like a spike. I think you should check if something like “Trim” is running when the system is idle—perhaps as part of automatic maintenance or something similar—and ask the Windows side to investigate which processes are performing I/O by capturing a performance log.

> It seems that the virtio drivers (0.1.271) crashes out,

They keep saying it doesn't show up anywhere or that it's crashing, so I'm not really sure what's going on...

By the way, this is the Proxmox forum.

 

[SteveITS]

271 should be the least problematic recent driver AFAIK.

re: TRIM, manually add this to the VM's /etc/pve/qemu-server/###.conf file for much faster TRIM: args: -global scsi-hd.discard_granularity=8388608
(edit: link - https://forum.proxmox.com/threads/ssd-optimize-not-working-inside-windows-2022-server.183978/)

Per various other threads what is the CPU in the VM set to? If host or some other recent CPU it may enable the virtualization features in Windows and slow everything down. See https://forum.proxmox.com/threads/h...sage-on-idle-with-windows-server-2025.163564/ and several other threads.

 

[Forgemaster98]

We have found that Oracle’s VirtIO drivers are currently the most stable option for server 2k25 (Oracle VirtIO Drivers for Microsoft Windows Version 2.3.2)