Time drift in Windows 2008 R2 64-bit SMP

[obrienmd]

I've been noticing pretty obvious time drift in Win2k8 R2 64-bit (also Win2k8 64-bit), especially when under i/o pressure. Happens w/ both IDE/e1000 and VirtIO/VirtIO driver sets.

I see the following RH info on this:
https://bugzilla.redhat.com/show_bug.cgi?id=577266
http://rhn.redhat.com/errata/RHBA-2010-0419.html

Host info:
Core i7 860 on P55 chipset
16GB DDR3
8 disk SAS RAID10 on Areca 1222

pveversion:
pve-manager: 1.5-9 (pve-manager/1.5/4728)
running kernel: 2.6.32-2-pve
proxmox-ve-2.6.32: 1.5-7
pve-kernel-2.6.32-2-pve: 2.6.32-7
pve-kernel-2.6.24-9-pve: 2.6.24-18
pve-kernel-2.6.24-8-pve: 2.6.24-16
qemu-server: 1.1-14
pve-firmware: 1.0-5
libpve-storage-perl: 1.0-13
vncterm: 0.9-2
vzctl: 3.0.23-1pve11
vzdump: 1.2-5
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.12.4-1
ksm-control-daemon: 1.0-3

 

[dietmar]

Please try to add the following to you VM configuration

args: -no-kvm-pit-reinjection

Does that help?

 

[obrienmd]

Yup, seems to have solved it. Perhaps this should be default for 64-bit Windows guests?

 

[obrienmd]

Err, looks to be back on the fritz - so I don't think this arg solved it. Trying rtc-td-hack arg now (as per http://www.linux-kvm.com/content/kvm-84-released-bug-fixes-and-qemu-merge).

 

[obrienmd]

And no luck. Things seem to equalize (or internet time sync just makes up for it), but seconds definitely don't 'tick' evenly on the system clock.

 

[dietmar]

Please can you post the VM config (/etc/qemu-server/<vmid>.conf)?

 

[obrienmd]

name: xyz.praece.com
bootdisk: virtio0
ostype: w2k8
memory: 2048
sockets: 2
virtio0: array:vm-112-disk-1,cache=none
vlan2: virtio=7A:02:4E:5A:73:FC
boot: cad
freeze: 0
cpuunits: 1000
acpi: 1
kvm: 1
args: -rtc-td-hack
ide0: none,media=cdrom

Have tried with -rtc-td-hack, -no-kvm-pit-reinjection, and nothing for args. Have tried with IDE and e1000 drivers instead of virtio. Nothing seems to help.

 

[Taliesin]

what file format for HDD do you use, us helped a lot converting qcow2 to raw

 

[obrienmd]

raw (lvm on hardware raid)

 

[tom]

test with one vCPU, any difference?

 

[obrienmd]

1 vCPU makes a big difference, there is minor clock instability but it is very close to real-time for the most part. However, we have 16 1.86ghz cores, 1 core on a fairly busy Windows 2008 R2 server makes things fairly slow

 

[tom]

can you give details about your CPU´s/Mainboard?

 

[tom]

forget it, I just read your first post.

 

[obrienmd]

Yeah, 3 different hosts showing these issues... The "bleed edge test" one in the first post:
Core i7-860
Biostar TP55

And our two "test production servers":
Intel S5000VSA each
Dual Xeon E5320 each

The two test production servers are the same hardware as our two production servers.

 

[tincboy]

Any luck on this issue?
I've a VM with 6 cores Windows 2008 R2 64 bits and I got 10 min drift back every two hour.

 

[tom]

pls open a new thread. include your 'pveversion -v' and also your VM settings. (cat /etc/qemu-server/VMID.conf)