On 11/08/2008, at 8:42 PM, Daniel P. Berrange wrote:
On Mon, Aug 11, 2008 at 09:18:19AM +0100, Richard W.M. Jones wrote:
On Mon, Aug 11, 2008 at 07:39:34PM +1200, james wrote:
This is what libvirt gives you (and lots more, eg. secure remote
access to hypervisors, bindings to Perl & many other languages,
etc.).
Can you be more specfic about what you couldn't do with libvirt?
I can give you such an example although I confess it could be due
to my
lack of understanding of the libvirt config. I have tried and
tried to
use libvirt to configure VMs within KVM using scsi disk images.
Usually
when tinkering/experimenting with RAID setups. It just will not
take it.
Starting a KVM based VM from the command line with the appropriate
settings and I have no problems. This inability to use scsi within
libvirt has been extremely frustrating until I took the plunge and
went
to to kvm command line.
However if you can point out an example xml config for a VM using
scsi
disk images that works then that would be very cool
Configuring SCSI disks with VMs in libvirt is no different to
configuring
any other kind of block based storage. The general description is
here:
http://libvirt.org/formatdomain.html#elementsDisks
Specifically though you'd want a disk section looking like
<disk type='block'>
<source file='/dev/sdf1'/>
<target dev='sda' bus='scsi'/>
</disk>
NB, there is no restriction on mapping to the target bus -ie a SCSI
disk in
the host can be mapped to a IDE disk in the guest, and vica-verca.
Also
note that the 'dev' attribute on the target isn't a guarenteed
device name
in the guest - it is merely used for ordering of devices when
spawning QEMU.
Now, the main fun you'll have is actually outside of libvirt -
namely that
on Linux SCSI disk names are not guarenteed stable across reboots. So
rather than using /dev/sdf1 you may want to consider one of the udev
created
stable paths under the directories /dev/disk/by-{id,path,uuid}, or
if you are
using a multipath enabled SAN, then a name under /dev/multipath/XXXX
Regards,
Daniel
To be clear I have not tried to use block devices but 5 image files as
I am experimenting. I have tried to create VMs using raw image files
under libvirt via the nice Virtual Machine Manager 0.5.3 and through
hand crafted xml files. Neither method will register the images as
scsi disks, it fails with an error. Getting the same disk images to be
regarded as scsi disks via the kvm command line is fine or if I fall
back to ide and limit. The ide test proves that the rest of the xml
file is correct.
Trying to define a domain using virsh using a disk section defined as:
<disk type='file' device='disk'>
<source file='/home/someuser/scsi-test/scsi-disk1.img'/>
<target dev='sda' bus='scsi'/>
</disk>
fails with this error:
libvir: QEMU error : Invalid harddisk device name: sda
Heck if I have missed the wood for the trees and there is a simple
correction to this definition that will make it work I would be one
happy camper.
Cheers.
-- James.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
