Re: [Xen-users] putting virtual domains into NFS mounts into turn mounted on DRBD devices.

[Javier Guerra <javier@xxxxxxxxxxx>]

On Thursday 20 April 2006 5:48 am, zHihaO wrote:
> Hi all,
>
> I'm worried about this because we are going to be using these virtual
> servers for webhosting. Email and web traffic will be quite high and I'm
> afraid its gonna thrash the harddisk with the high disk i/o expected.
> And if the disk for storing the domUs are NFS exports..I'm not so sure
> if performance will be hit badly. comments anyone? or has anyone got a
> better suggestion??

if you're using DRBD, you'll get a shared block device.  then you're planning 
to add tree layers over that shared block device: 1) a filesystem.  2) NFS. 
3) image files.

instead of that, you could put LVM (or clvm) over that shared block device. it 
would be far less overhead

and, if i understand correctly, you have two boxes, each one with a big local 
disk, currently using LVM.  that doesn't let you do live migration, but get's 
the best performance, right?

if you're not planning to add a real SAN (iSCSI or otherwise), or more boxes 
in the near future, i think your best configuration would be something like 
that:  use DRBD to replicate the PV between the two boxes, and then use LVM 
to create several LVs over that PV.  the disk access would be mainly local, 
only the DRBD replication would happen transparently and asynchronously, 
without impacting disk access time.

-- 
Javier

pgp705GP4j5t2.pgp
Description: PGP signature

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users