[fedora-virt] Directory Passthrough

[Robert Locke]

On Mon, 2014-03-17 at 11:22 +0000, Richard W.M. Jones wrote:
> [A shared filesystem sounds like what you want, but given that qemu
> only has 9pfs and the KVM team have deemed that this is unsupportable,
> read on ...]
> 
> On Sun, Mar 16, 2014 at 07:17:00PM -0400, Robert Locke wrote:
> > Yeah, NFS mounting has been a little problematic/inconsistent for me.
> > 
> > Specifics: 
> > 
> > 1) My "/content" directory on the hypervisor host has both direct data
> > but also a loop mount of an iso. The iso and its mountpoint both reside
> > inside /content.
> 
> You could export the ISO readonly to multiple guests as a block
> device.  Put this into the libvirt XML:
> 
>     <disk type='block' device='cdrom'>
>       <driver name='qemu' type='raw'/>
>       <source file='/path/to/cd.iso' />
>       <target dev='sdc' bus='virtio-scsi' tray='open'/>
>       <readonly/>
>       <shareable/>
>     </disk>
> 
> How often does the non-ISO content change?  If it's essentially static
> over long periods you could make it into an ISO or squashfs (or
> virt-make-fs) and export it to the guests.  However this becomes quite
> clumsy if the content changes much.
> 
<snip>

Hmmmm... You know I need to take a closer look at all these libguestfs
tools......

virt-make-fs is interesting to me.

There is a Monday morning setup and the information in /content doesn't
change after that setup.

I'm sure I can experiment myself, but I wonder how long virt-make-fs
takes to create, say, a qcow2 copy of the tree of, say, 12GB?  This
might be a way around my NFS problems....

--Rob