[DRBD-user] DRBD, Xen, HVM and live migration

Arnold Krille arnold at arnoldarts.de
Fri Sep 23 12:49:36 CEST 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Friday 23 September 2011 12:29:08 Felix Frank wrote:
> On 09/23/2011 11:09 AM, Bart Coninckx wrote:
> > What would be a good strategy these days to accomplish this on a two
> > node system (no external storage)?
> 
> A simple LVM -> DRBD -> Guest Partitions approach has worked fine for me.
> 
> > Do I need to complexity of a clustered filesystem?
> 
> Where would that come into play? Are you thinking of using disk image
> files for your guests? That sounds icky.

Well, a two-node cluster using (clustered-)lvm to provide the images doesn't 
really extend to three or four nodes as far as I understand it.

Thats why we started doing two nodes, drbd dual-primary, gfs, pacemaker and 
virtualization to run virtual machines (and provide other services) in ha-
mode. That worked so good my collegue had real problems killing a machine 
without knowing pacemaker...

And to extend this, the two master nodes will export the drbd-masters via 
iscsi and the additional nodes will bind the iscsi and run pacemaker+gfs+virt 
and be able to run the same virtual machines. But as this is only planned for 
the near future, if anyone has any good or bad experience with this and is 
willing to share, I would be glad.

And if there is a penalty for using image-files for the vm, I haven't 
encountered any. Reading and writing is about the same regardless whether its 
the local disk, the drbd-resource or the image-file inside the vm.

Have fun,

Arnold
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20110923/88650f17/attachment.pgp>


More information about the drbd-user mailing list