Re: [Libvir] [PATCH] Linux-VServer support
Daniel P. Berrange wrote:
On Wed, Jan 09, 2008 at 10:51:20PM +0100, Daniel Hokka Zakrisson
wrote:
> This patch implements support for Linux-VServer guests. It is currently
> missing vcpu and console support, and the necessary virsh code to
> support
> it, but is otherwise pretty feature complete.
>
> This is an XML dump from one of my guests:
> <domain type='vserver' id='40010'>
> <name>lenny</name>
> <uuid>19e12957-261a-5a06-d6d0-89917d6d439f</uuid>
> <memory>2048000</memory>
> <os>
> <hostname>lenny.test</hostname>
> <type arch='i686'>vserver</type>
> </os>
> <devices>
> <interface type='ethernet'>
> <ip prefix='24' interface='dummy0'
address='192.168.20.4' />
> </interface>
Minor nitpick - if dummy0 is the device associated with the
container in the host's context, then it should be
<interface type='ethernet'>
<ip prefix='24' address='192.168.20.4' />
<target dev="dummy0"/>
</interface>
> <interface type='ethernet'>
> <ip prefix='24' interface='dummy0' type='range'
> address='192.168.32.100' address2='192.168.32.200'/>
> </interface>
> <disk type='directory' device='directory'>
> <source directory='/vservers/lenny' type='auto'
> options='defaults'/>
> <target directory='/'/>
> </disk>
I think I'd prefer to call these '<filesystem>' elements instead
of '<disk>' because they don't really have the same functional
properties as disks.
Overloading <disk> was done mostly to keep in line with the other
backends. I'd rather just add some logic to attach-disk than an entirely
separate attach-filesystem...
Are the sources actual directories too ? Or are the sources block
devs / files which are mounted inside ? The example paths suggest
the former, but the presence of filesystem type info suggests the
latter.
Either. That's what the <disk type='directory'> specifies, and the
point
of the second example using type='block'. type='file' is also supported,
when used in conjuction with options='defaults,loop'.
Dan.
--
Daniel Hokka Zakrisson