Windows Server 2022 VM freezed - virtio: zero sized buffers are not allowed

[Max2048]

Konnte schon jemand testen ob die Parameter in der VM-Config zum Deaktivieren der Hotplug-Funktionalität helfen auf einer q35-Machine?

Ich werde heute einen Test damit starten. Mein letzter Test mit dem Zurückstellen des Sprachpakets auf Englisch scheint das Problem nicht vollständig zu beheben, nach 2 Wochen Uptime gab es nun erneut einen Crash am Wochenende mit den selben Symptomen wie im ersten Post beschrieben. Dennoch deutlich weniger Abstürze als mit deutschem Sprachpaket (alle 2-3 Tage).

 

[davemcl]

Did you ever get a resolution to this?
This morning I got the same on the host

QEMU[2174]: kvm: virtio: zero sized buffers are not allowed

In the Windows/SQL VM it follows with the following and the VM has to be reset

Reset to device, \Device\RaidPort4, was issued.

 

[Max2048]

@davemcl Unfortunately no, I added a weekly automatic restart via scheduled tasks.

 

[davemcl]

Thanks @Max2048

What kernel version are your Proxmox hosts?

 

[Max2048]

Thanks @Max2048

What kernel version are your Proxmox hosts?

5.15.64-1-pve

 

[Max2048]

Thanks @Max2048

What kernel version are your Proxmox hosts?

@davemcl I applied some updates and my VMs are running fine so far. I disabled the scheduled restart in my VM. I found a possible fix in the pve-qemu-kvm package: https://git.proxmox.com/?p=pve-qemu.git;a=commit;h=d5f6ef56f0226b0efe34d6e08bfbad1706da3326

What qemu version are you using? apt list --installed | grep pve-qemu-kvm
Make sure to run at least 7.2.0-4 of that package. Do you still have these crashes with this version or above?

 

[davemcl]

Hey @Max2048
Errors alternate between
QEMU[2198]: kvm: Desc next is 3
and
QEMU[1961]: kvm: virtio: zero sized buffers are not allowed

Im on a later version
pve-qemu-kvm/stable,now 7.2.0-8 amd64 [installed]

I may try upgrading a single node to the 8.0 Beta, failing that I'll have to look at a different solution.

 

[Max2048]

@davemcl Interesting, I've never had the first error. Maybe we're actually having different issues. No more crashes for me so far (2 weeks) with pve-qemu-kvm 7.2.0-8.

Did you try updating the virtio drivers in the VMs?

 

[davemcl]

Ive now managed to upgrade the cluster to PVE8, will report back if that helps this issue.

 

[davemcl]

So I got about 5 days past the PVE 8 upgrade and got the less frequent error of
QEMU[4927]: kvm: Desc next is 3

Since the beginning of June Ive seen "Desc next is 3" error twice, I usually get this one more often (6 times)
QEMU[1884]: kvm: virtio: zero sized buffers are not allowed

Result is same though, VM cant be shutdown & has to be killed.
VM just shows the fairly generic error
Reset to device, \Device\RaidPort5, was issued.