> thank your for your suggestion.
> I am not familiar with nbd, I was hoping you could help me out a little
> can you tell me what the difference is between enbd and gnbd? which
> is better?
I'm afraid I don't know offhand.
> does xen support gnbd?
Quite possibly; but if it doesn't support it already it's probably only a
small hack to make it work.
> from what i read about nbd it seems to export whole partitions, but
> can it export folders like nfs or smb?
It exports block devices, so it's below the level of exporting folders.
> is nbd able to export a loop back filesystem?
Yes, but you wouldn't need to bind the file to the loop device. Just give nbd
the file and tell it to export it directly, thus avoiding the problems of
using loop devices.
> On 8/13/07, Mark Williamson <mark.williamson@xxxxxxxxxxxx> wrote:
> > > How safe is running a domU sda1 off of an image file located on an NFS?
> > If you can use it as a tap:aio it should work OK. Don't use it as a
> > file: disk (unless it's an HVM guest, in which case it's probably OK - I
> > apologise but the disk handling is a bit weird at the moment!).
> > The reason for this advice is that file: VBDs for paravirt guests are
> > implemented using losetup to bind the file to the loop device. Doing
> > this on an NFS mounted filesystem is known to cause nasty memory usage
> > problems.
> > Hopefully even if this doesn't make sense to you right now, you'll see
> > what I mean when you've got a bit further in setting stuff up ;-) If
> > not, folks on here should be able to fill you in.
> > > I am thinking about setting up a WEB server for hosting our internal
> > > web apps on
> > > a domU. I am thinking about creating the root disk image on the NFS
> > > folder so that when i do i live migration the disk image is available
> > > on the new server.
> > >
> > > is this safe?
> > Should be fairly safe for HVM guests, or for paravirt guests using
> > tap:aio...
> > I'm not sure whether the semantics of NFS will give you guaranteed data
> > persistence (e.g. after a sync in the guest will the writes really have
> > hit the disk at the server).
> > There used to be some race conditions with respect to storage when live
> > migrating a guest, but they generally didn't seem to break things for
> > anybody, and I think they may be fixed now anyhow.
> > Sorry for being a bit vague, I'm not so familiar with the specifics in
> > this area at the moment.
> > > how is the performance?
> > > is there a better way to do this without investing in SAN?
> > Doing this over NFS is probably not optimal. You could look at using
> > ENDB (the Network Block Device). The Xen config file format allows for
> > remote NBD disks to be specified directly in the config file so that live
> > migrations will automatically work properly, etc.
> > You need a separate machine (or virtual machine) to run as the NBD
> > server, because it's not safe to mount an NBD server on the same machine
> > it's running on. Maybe you could run the NBD server on the storage
> > server for your network, if it has the capacity to cope with the virtual
> > machine disk traffic? It's a bit less powerful than iSCSI but simpler to
> > set up and it gives you similar "SAN over ethernet" functionality.
> > Hope that helps,
> > Cheers,
> > Mark
> > --
> > Dave: Just a question. What use is a unicyle with no seat? And no
> > pedals! Mark: To answer a question with a question: What use is a
> > skateboard? Dave: Skateboards have wheels.
> > Mark: My wheel has a wheel!
Dave: Just a question. What use is a unicyle with no seat? And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!
Xen-users mailing list