W2025 virtio NIC -> connection drop outs

P

pme242

New Member
Oct 19, 2025
5
0
1
Hi,

I have network issues with the virtio NIC on Windows Server 2025. The network connection drops for some seconds every 1-2h

PVE: 9.0.11 (fully patched)
virtio: virtio-win-0.1.285-1
VM: Windows Server 2025 (setup like suggested by Proxmox)

I've switched the vNIC to Intel E1000 - all good.

Any suggestions?

Thanks in advance.

PM
 
YES :)

and for info th VM

agent: 1,fstrim_cloned_disks=1
balloon: 0
bios: ovmf
boot: order=sata1;sata0
cores: 4
cpu: host
efidisk0: SSDTHIN1:vm-401-disk-0,size=4M
machine: pc-q35-10.0+pve1
memory: 20480
meta: creation-qemu=9.2.0,ctime=1747982103
name: WS-NAS1
net0: virtio=BC:24:11:69:6A:26,bridge=vmbr0,queues=4
numa: 1
onboot: 1
ostype: win11
parent: OK
sata0: none,media=cdrom
sata1: SSDTHIN1:vm-401-disk-1,cache=writeback,discard=on,size=90G,ssd=1
scsihw: virtio-scsi-single
smbios1: uuid=564d1878-18c9-8aa5-c7f0-b8e3bbccd236
sockets: 1
startup: order=2,up=1000
tags: windowsserver
tpmstate0: SSDTHIN1:vm-401-disk-2,size=4M,version=v2.0
vga: std
vmgenid: dc440efe-aed3-42e2-9f48-e577c84a447f

[OK]
#vers nvx 2.8.5
agent: 1,fstrim_cloned_disks=1
balloon: 8192
bios: ovmf
boot: order=sata1;sata0
cores: 4
cpu: host
efidisk0: SSDTHIN1:vm-401-disk-0,size=4M
machine: pc-q35-9.2+pve1,viommu=virtio
memory: 20480
meta: creation-qemu=9.2.0,ctime=1747982103
name: WS-NAS1
net0: virtio=BC:24:11:69:6A:26,bridge=vmbr0,queues=4
numa: 1
onboot: 1
ostype: win11
running-nets-host-mtu: net0=9000
runningcpu: host,hv_ipi,hv_relaxed,hv_reset,hv_runtime,hv_spinlocks=0x1fff,hv_stimer,hv_synic,hv_time,hv_vapic,hv_vpindex,+kvm_pv_eoi,+kvm_pv_unhalt
runningmachine: pc-q35-9.2+pve1
sata0: none,media=cdrom
sata1: SSDTHIN1:vm-401-disk-1,cache=writeback,discard=on,size=90G,ssd=1
scsihw: virtio-scsi-single
smbios1: uuid=564d1878-18c9-8aa5-c7f0-b8e3bbccd236
snaptime: 1760896684
sockets: 1
startup: order=2,up=1000
tags: windowsserver
tpmstate0: SSDTHIN1:vm-401-disk-2,size=4M,version=v2.0
vga: std
vmgenid: 2efc3f19-939e-4fbb-97f4-dfd877ce9807
vmstate: SSDTHIN1:vm-401-state-OK
 
I've reverted to 1.271, but still having some outages - less than before.
So I've switched to E1000 to fix this issue temporary.

Any idea where to report this issue?
 
In my case this is a ws2025 with iscsi to truenas (on the same proxmox)
No issue with network since reverting (truenas is sensible to network issues with iscsi)
 
Ran into the same issue running Windows Server 2022 on a Veeam proxy between our SAN and the backup repository with VirtIO virtual hardware used everywhere possible. Backups were failing and iscsi/disk errors were filling the System log. Tried various reconfigurations of the VM with no impact to the errors.

Upon reverting the networking drivers to 100.100.104.27100 the errors stopped and backups began finishing as expected. I strongly suggest pulling the 0.1.285 release of the VirtIO Windows drivers from the Proxmox wiki.
 
Last edited:
Do the servers without the errors use iscsi connections? If so, do they experience multi-gigabit load?
 
Perhaps the error is specific to iscsi connections. Reviewing logs of another Windows 2022 VM with VirtIO on 285 but no iscsi, multi-gig traffic does not seem to experience any errors.
 
Unfortunately also the E1000 drops sometimes.
How to report this issue?
Who's in charge for the the fix?

no iscsi - just a webservice
 

Attachments

  • Screenshot 2025-10-23 142654.png
    Screenshot 2025-10-23 142654.png
    56.7 KB · Views: 4
Last edited:
Summary:
When using VirtIO driver version 0.1.285, the Windows iSCSI Initiator immediately drops the connection during any disk operation. Rolling back to version 0.1.271 completely resolves the issue.

Environment:
  • Host: TrueNAS (serving iSCSI target)
  • Client: Windows Server 2022
  • Use case: Veeam Backup repository connected via iSCSI
  • Network: Dedicated iSCSI interface, no VLANs or jumbo frames
  • MPIO: Disabled
Observed behavior:
After connecting the iSCSI target and initiating a write or format operation, the connection drops within a few seconds. Windows Event Log records a series of errors from the iScsiPrt and disk providers, including:
  • “Target sent an iSCSI PDU with an invalid opcode. Dump data contains the entire iSCSI header.”
  • “The initiator could not send an iSCSI PDU. Error status is given in the dump data.”
  • “Disk 1 has been surprise removed.”
The issue reproduces consistently with VirtIO 0.1.285 and does not occur with VirtIO 0.1.271 under identical conditions.

Analysis:
The invalid opcode errors suggest malformed or corrupted iSCSI PDUs being sent by the initiator. This points to a regression in the VirtIO SCSI/TCP handling layer introduced in 0.1.285, likely related to queue management or buffer alignment changes.

Workaround:
Revert the VirtIO driver to version 0.1.271 — this version maintains stable iSCSI connectivity.
 
  • Like
Reactions: need2gcm
MZoltan said:
Summary:
When using VirtIO driver version 0.1.285, the Windows iSCSI Initiator immediately drops the connection during any disk operation. Rolling back to version 0.1.271 completely resolves the issue.

Environment:
  • Host: TrueNAS (serving iSCSI target)
  • Client: Windows Server 2022
  • Use case: Veeam Backup repository connected via iSCSI
  • Network: Dedicated iSCSI interface, no VLANs or jumbo frames
  • MPIO: Disabled
Observed behavior:
After connecting the iSCSI target and initiating a write or format operation, the connection drops within a few seconds. Windows Event Log records a series of errors from the iScsiPrt and disk providers, including:
  • “Target sent an iSCSI PDU with an invalid opcode. Dump data contains the entire iSCSI header.”
  • “The initiator could not send an iSCSI PDU. Error status is given in the dump data.”
  • “Disk 1 has been surprise removed.”
The issue reproduces consistently with VirtIO 0.1.285 and does not occur with VirtIO 0.1.271 under identical conditions.

Analysis:
The invalid opcode errors suggest malformed or corrupted iSCSI PDUs being sent by the initiator. This points to a regression in the VirtIO SCSI/TCP handling layer introduced in 0.1.285, likely related to queue management or buffer alignment changes.

Workaround:
Revert the VirtIO driver to version 0.1.271 — this version maintains stable iSCSI connectivity.
Click to expand...
same with WS 2025
 
You must log in or register to reply here.
Top Bottom