Hello!

I changed my system from [Asrock Rack EPC612D8A & Xeon E5-2690 v4] to [Gigabyte MC62-G40 & Threadripper PRO 3945WX] and wanted to change my cache pool from [2x Samsung 850 EVO 250GB] to [Samsung NVME 970 EVO Plus 1TB].

The first part worked fine, but when I put the new board into the old case I remembered that I forgot to move the files from the cache to the array and had to put the old board in a different case because the new one hadn't enough SATA ports. That worked more or less fine.

I then put my NVME drive into the new board and started it without the unraid USB stick and used the OS on it to migrate my data from the NVME to a new [Samsung 990 EVO Plus 4TB].

Then I put in the unraid USB stick again and replaced the two Sata SSDs with the NVME 1TB drive and started the array. I then saw that the previous OS and data was still on it, so I manually removed the files on it in the unraid GUI. Then I moved appdata, system and domains back on the new NVME 1TB which worked seemingly fine. But when I started the VM manager and Docker, and tried to bind my new 4TB NVME to my old VM (and change other things like core counts) it throw me an error that it couldn't write into libvirt.

So I restarted and moved the files from the NVME back to the array and tried to format the NVME, moved the files back on it and tried again with the same error, I tried that a few more times and last I formated it with btrfs, moved the files on it again and now it is completely unavaible with the data on it and the following error:

BTRFS error (device nvme1n1p1): bdev /dev/nvme1n1p1 errs: wr 1, rd 0, flush 0, corrupt 0, gen 0

Cache shows:

Unmountable: unsupported or no filesystem

To summarize my hardware changes:

From

Asrock Rack EPC612D8A

Xeon E5-2690 v4

8x16GB (128 GB) Ram

Samsung NVME 970 EVO Plus 1TB (exclusively bound to VM)

Cache: 2x Samsung 850 EVO Plus

Several HDDs

To

Gigabyte MC62-G40 [https://www.gigabyte.com/de/Enterprise/Server-Motherboard/MC62-G40-rev-1x#Support-Manual]

Threadripper PRO 3945WX [https://www.amd.com/en/support/downloads/drivers.html/processors/ryzen-threadripper-pro/ryzen-threadripper-pro-3000wx-series/amd-ryzen-threadripper-pro-3945wx.html]

same 8x16GB Samsung M393A2G40EB1-CRC [https://semiconductor.samsung.com/dram/module/rdimm/m393a2g40eb1-crc/]

Samsung 990 EVO Plus 4TB (to be bound exclusively to VM)

Cache: Samsung NVME 970 EVO Plus 1TB

Several HDDs

Things I tried:

• changed in the bios "Power supply idle control" to "typical current idle"

• tried "tried btrfs rescue zero-log /dev/nvme1n1p1

  ERROR: mount check: cannot open /dev/nvme1n1p1: No such file or directory

  ERROR: could not check mount status: No such file or directory"

waterline-sv-diagnostics-20251117-1916.zip

Post the output from btrfs fi show

After starting up unraid again, the NVME seems to work like normal. I have not yet started VM manager or docker. Now that I have access to the files again, should I copy them to the array again and format the NVME again, but different. If yes, how should I format it?

root@Waterline-SV:~# btrfs fi show

Label: none uuid: d0446c4c-64c2-4ce1-ae7d-24ec07404bb2

Total devices 1 FS bytes used 331.20GiB

devid 1 size 931.51GiB used 338.02GiB path /dev/nvme1n1p1

Post new diags pelase.

Okay, after starting VM manager again and trying to edit my VM:

Failed to create file '/etc/libvirt/qemu/Windows 11.xml.new': Read-only file system

waterline-sv-diagnostics-20251118-2009.zip

The NVMe device is being passed through to a VM, so when the VM starts, Unraid loses the device; correct that.

Which I tried:

Failed to create file '/etc/libvirt/qemu/Windows 11.xml.new': Read-only file system

Creating a new VM:

operation failed: unable to find any master var store for loader: /usr/share/qemu/ovmf-x64/OVMF_CODE-pure-efi-tpm.fd

You are meant to passthrough the NVME by making a check mark in front of it in 'system devices' and the thing marked there is the new NVME(4TB), the old one (1TB) is not checked?

waterline-sv-diagnostics-20251118-2037.zip

Edit:

So, I have removed all bindings, restarted, and bound the 4TB NVME again. It seems to work for the VM manager now, but I haven't yet moved the appdata back.

Anyway, my passthrough GPU doesn't show video anymore and also the sound card part of the GPU only shows Nvidia device, the graphic card on the other hand shows normally. Should I made a new thread for that problem or continue this one?

waterline-sv-diagnostics-20251119-0327.zip

Edited November 19, 2025Nov 19 by blkopler

11 hours ago, blkopler said:

So, I have removed all bindings,

You only needed to remove the NVMe device, but pass through the GPU again.

On 11/19/2025 at 8:29 AM, JorgeB said:

You only needed to remove the NVMe device, but pass through the GPU again.

I never had bound the GPU on the old system, it never seemed to be necessary.

All seems to work, I had to change a few options (SR-IOV Support - enable; Resizable Bar - enable)in the bios and to bind the GPU to make the GPU usable in the VM.