11. July 2013

W2k8 in KVM is not able to reach network

I found interesting issue while testing W2k8 with KVM. I used bridged network with br0 interface.

<interface type='bridge'>
 <mac address='52:54:00:...'/>
 <source bridge='br0'/>
 <model type='e1000' />
</interface>

Windows server was not able to reach network. Network card was on-line, but ping was not working and network was not reachable.

Then I found PilotJLR’s solution. I just removed line with model type=’e1000′:

<interface type='bridge'>
 <mac address='52:54:00:...'/>
 <source bridge='br0'/>
</interface>

Then I redefined virtual machine and start it again:

virsh define virtual.xml
virsh start virtual

29. May 2013

Solution: KVM guest IO very slow

One problem of virtual server based on qemu KVM is that IO operations are slow in some cases.

The most common reason is that virtual disks are stored on RAID and virtual is using default HDD configuration.

It is recommended to turn off cache and set io operations to native:

<driver name='qemu' type='raw' cache='none' io='native'/>

There is very good article at serverfault. You can find there further explanations.

31. December 2012

KVM error: internal error no HVM domain loader – solution

It’s very pleasant situation when you reboot Linux server and it starts.

It’s very unpleasant situation when this server contains many KVM virtual servers and none of them start. 🙁

virsh list displays just this:

Id Name State
----------------------------------
 0 Domain-0 running

That’s nonsense. There was no such virtual with name Domain-0.

When you try to create domain you can see this fancy error message:

[root@server qemu]# virsh create ./virtual-01.xml
error: Failed to create domain from ./virtual-01.xml
error: internal error no HVM domain loader

What went wrong? How to fix it?

I spare you long story with server diagnostic.

The problem was that server started different kernel after reboot. In this case it was different kernel with Xen support. Centos kernel with Xen support does not have KVM support. That’s quite logical. It was sufficient to change default kernel in grub.conf to non-Xen version, restart machine and everything was working like a charm.