Trouble to resize Passthrough Physical Disk to Virtual Machine

felie

New Member
Jul 17, 2025
7
1
3
Hi
for large partitions located in iSCSI SAN (~100To), I use Passthrough Physical Disk to Virtual Machine and the "qm set" command to add the iSCSI LUN as new virtual SCSI disk to the VM.
But when I resize the SAN iSCSI volume, I see that /dev/disk size is upgrade on proxmox server, but upgrade information seem to not be forwarded to the virtual machine (noting in kernel message after scsi rescan).
I hope it would work fine if I change the disk scsi size in the `vm`.conf file and reboot de VM, but becaus the VM is a data nfs server used by several VM for intensive computing, I need to resize the Passthrough Physical Disk without reboot the VM.
Do you have any suggestion, please ?
 
I did not resolve the trouble but I found a workaround : I migrated the VM to an other PVE node and migrate back. I also notice that for Passthrough Physical Disk, the disk size indicated in the `VMID`.conf file is not used.
If you have an idea howto to refresh QEMU passthrough HARDDISK informations sent to the VM without reboot or migrate it, I am interested.
 
Just to clarify: You did a scsi bus rescan and scsi disk rescan on the host and the guest?

EDIT: typo
 
Last edited:
Just to clarify: You did a scsi bus rescan and scsi disk rescan on the host and the guest?

EDIT: typo
I think so. I performed on PVE and VM the command "echo "- - -" > /sys/class/scsi_host/hostX/scan" and "echo 1 > /sys/class/scsi_device/X/device/rescan"
Something else to do ?
 
Yes, that should do the trick. I normally use /sys/class/scsi_disk, but that's just cosmetics and the same file in the backend.

You could try to look at the info via the qemu monitor (accessible via the PVE UI) and run an info block and check if the data is updated.

Another approach would be to access the iscsi LUN directly from the guest. With this, you omit the passthrough step altogether.
 
Hi, the "info block" do not indicate volume size. I was hoping to use block_resize command but it seems to be only for image files (raw, qcow2, etc) but not for iSCSI LUNs

I heard that it is better tu manage iSCSI with host rather than VM with HA and PVE clustering. I do not know why, but I trusted
 
Last edited: