Windows KVM frequent restarts

[Sakis]

A KVM with windows 2008 R2 has frequent unwanted reboots.

I noticed that reboots happen when cpu load starts to grow.

bootdisk: ide0
cores: 4
cpu: qemu64
ide2: none,media=cdrom
memory: 16384
net0: e1000=8E:52:D9:AD:33:E5,bridge=vmbr0,rate=125
onboot: 1
ostype: win7
sockets: 1
virtio0: A:110/vm-110-disk-1.raw,format=raw,size=80G
virtio1: B:110/vm-110-disk-1.raw,format=raw,backup=no,size=500G

proxmox-ve-2.6.32: 3.2-126 (running kernel: 2.6.32-29-pve)
pve-manager: 3.2-4 (running version: 3.2-4/e24a91c1)
pve-kernel-2.6.32-29-pve: 2.6.32-126
lvm2: 2.02.98-pve4
clvm: 2.02.98-pve4
corosync-pve: 1.4.5-1
openais-pve: 1.1.4-3
libqb0: 0.11.1-2
redhat-cluster-pve: 3.2.0-2
resource-agents-pve: 3.9.2-4
fence-agents-pve: 4.0.5-1
pve-cluster: 3.0-12
qemu-server: 3.1-16
pve-firmware: 1.1-3
libpve-common-perl: 3.0-18
libpve-access-control: 3.0-11
libpve-storage-perl: 3.0-19
pve-libspice-server1: 0.12.4-3
vncterm: 1.1-6
vzctl: 4.0-1pve5
vzprocps: 2.0.11-2
vzquota: 3.1-2
pve-qemu-kvm: 1.7-8
ksm-control-daemon: 1.1-1
glusterfs-client: 3.4.2-1

Event log doesnt seem to help.

 

[tom]

could have many reasons, also inside windows.

windows VMs are known to run stable on proxmox. check the windows dump file.

 

[spirit]

Hi,

It's possible that the problem is cpu reclocking or dynamic cores enable/disable when cpu usage grow up.
Check in your server bios, set gouvernor to maximum performance, disable c1e, disable c-states. (and maybe other option for dynamic cpu freq, dynamic cores, turbo boost, ...)

The problem is that this cause clock problems, and windows guest don't like that and crash.

 

[remark]

I have same problem with Windows Server 2008 (not R2). Random BSODs with different error codes (0x3E, 0x101, 0x1E etc.)
Tried IDE and VirtIO as disk controller, Intel and VirtIO as network card.
C-states are disabled in server BIOS.

 

[Sakis]

Investigating the dump files showed me memory related problems.

All the dumps from the restarts had bug check string = PAGE_FAULT_IN_NONPAGED_AREA caused by win32k.sys driver.

Kvm.conf has fixed memory at 16384MB

[FONT=monospace]Type 'help' for help.
# info balloon
balloon: actual=16384 max_mem=16384[/FONT]

i had installed virtio balloon driver but now i see that i have not blnsvr service installed.

Should i try disabling virtio balloon driver and see the results or installing the blnsvr maybe?

I dont mind having graphs or full memory allocation on the host. I want stability.

Thanks

 

[holgihero]

I have same problem with Windows Server 2008 (not R2). Random BSODs with different error codes (0x3E, 0x101, 0x1E etc.)Tried IDE and VirtIO as disk controller, Intel and VirtIO as network card.C-states are disabled in server BIOS.

We also run one Server 2008 (based on Vista), which we found unstable mainly with VirtIO (crashing weekly). We switched to IDE and Intel NIC, tried a lot of things. Now it crashes only twice a year but we can not get it a hundred percent stable. 2008 R2 is much better (and slower).We explained to the customer, he accepts...Regards, Holgi

 

[remark]

Okay, I think the same way.
Thank you.

 

[e100]

This:

It's possible that the problem is cpu reclocking or dynamic cores enable/disable when cpu usage grow up.
Check in your server bios, set gouvernor to maximum performance, disable c1e, disable c-states. (and maybe other option for dynamic cpu freq, dynamic cores, turbo boost, ...)

The problem is that this cause clock problems, and windows guest don't like that and crash.

In addition to BIOS Settings I have found some kernel command line options that increase performance too.
DISCLAIMER, use this information at your own risk, if you melt your CPU you only have yourself to blame.
I have some Sandy Bridge E5-2650 2.0Ghz CPUs
Added this to /etc/default/grub ran update-grub then rebooted:

GRUB_CMDLINE_LINUX="intel_idle.max_cstate=0 processor.max_cstate=1"

All 8 cores now run at 2.4Ghz, according to the i7z utility I used to check.

With same option an E5-2690 2.9Ghz all cores run at 3.3Ghz
This allows you to take advantage of Turbo Boost without the clock speeds changing all the time.

2008 R2 is much better (and slower)

Maybe this will help: Control Panel -> Hardware -> Power Options -> Select High Performance
Made a huge difference for me.

I've been running Windows 2003, 2008, 2008 R2 and 2012 R2 all without issues.
I use virtIO for disks and e1000 for network on all Windows VMs.

The most common issues I have had are hardware problems, in this order:
RAM, Disks(Firmware update fixed), RAID Cards (Firmware fixed), Power supplies, and most recently a defective memory controller in an E5-2690 CPU.
Whenever I see new random issues RAM is the first suspect and memtest has often NOT found a problem when there was indeed defective RAM.

 

[spirit]

virtio balloon driver but now i see that i have not blnsvr service installed.

Should i try disabling virtio balloon driver and see the results or installing the blnsvr maybe?

balloon:0 to disable balloon device.

for blnsrv install :
https://pve.proxmox.com/wiki/Dynamic_Memory_Management

 

[remark]

BSODs gone.
I have several Windows Server VMs (2008, 2008R2, 2012R2) on my Proxmox Server.
Windows Server 2008 system crashed with BSOD once or twice in two weeks. Windows Server 2008R2 and 2012R2 was stable, I've got never BSODs with them.
Recently I've got a new Supermicro server with 2 x Xeon E2650v2 CPUs and 256 Gb RAM, SAS RAID with BBU, and built a 2-node cluster, without shared storage for now.
When I moved Windows Server 2008 VM to new server, BSODs began to happend more frequently, even 2-3 times per day.
I have upgraded Windows Server 2008 to Windows Server 2008R2 and BSODs seems to be gone away. Other Windows Server machines (2008R2 clean installed, 2012R2) also works stable.
Windows Server 2008 had BSODs with both IDE or VirtIO (different versions) as disk controller and both Intel E1000 and VirtIO as network controller.
So I guess this is Windows Vista kernel fault, which Windows Server 2008 built on.

P. S. All Windows Server systems is x86_64, of course.