Proxmox 8 crashes when copying data inside windows VM

[xoffio]

Hi,
I have a group of Precision 3680 machines with Intel(R) Core(TM) i9-14900K processors.

I am creating only one Windows 10 VM on each Proxmox machine, and I am using pass-through for the disk. However, after copying a large amount of data (In the VM) to the disk, Proxmox completely freezes.

I have attached the output of journalctl when Proxmox freezes (see journalctl_crash.txt).

This issue has only occurred on 2 out of the 6 machines so far.

I noticed that in the machines without issues, the SSDs all share the same IOMMU group. In the other hand, the machines with issues have SSDs in separate IOMMU groups. Here are the screenshots for reference:




I have tried using LVM, Directory, and ZFS instead of passing through the disk, but Proxmox hangs as well.
I read forums about people with similar problems. They suggested it might be ZFS, but I was not using ZFS. Even though I was not using ZFS, I reduced the RAM and cores of the VM, but that didn't solve the problem.

I have also tried the following:
- adding `clearcpuid=600` to `GRUB_CMDLINE_LINUX_DEFAULT`
- adding `mitigations=off` to `GRUB_CMDLINE_LINUX_DEFAULT`
- installing intel-microcode
Unfortunately, none of these solutions have worked.

I appreciate any help.

 

[tld8102]

Same, windows VM crashed... not even blue screen, just if the Proxmox reset button was pressed. except I'm copying 1.8TB via file explorer from a network drive trueness to a virt-io scsi 3-ish TB virtual drive (mirrored zfs). All of this runs on the same machine and sometimes affects the Truenas VM too. although there's no error reported usually seen there is an "unexpected shutdown".

edit: the windows VM doesn't seem to 'reset' when the virtual drive is set to IDE
edit2: I suspect it might have something to do with the HBA. now my truenas VM is just 'stopping' when copying to a test windows VM. the only difference is that it didn't have the 3.5TB virtual drive mounted. My challenge now is that I somehow need to find a way to passthrough the HBA card but it's connected via the chipset and shares the same IOMMU group. Because that same HBA card has 6 of 8 of the hard drives passed through to truenas while the other two are directly exposed to Proxmox as a zfs mirror. In the past Truenas has been rock solid when the last 2 drives were passed through. Even still now I've detached the the virtual drive drive from the VM and Truenas is starting to crash. I can only copy roughly about 25GB before trueness crashes, I strongly suspect my HBA has gone bust OR something to do with the drivers. like trueness and windows don't show a unexpected shutdown message or anything. more testing needed