[DRBD-user] cluster configuration question

Graham Clinch g.clinch at lancaster.ac.uk
Fri Jul 7 16:46:44 CEST 2006

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


Hi Miles,

> I'm wrestling with possible configurations for a 2-node, high 
> availability cluster.
> 
> What I seem to be zeroing in on is:
> - using Xen to provide several virtual servers on each box
> - using md to raid together the disks on each server
> - using DRBD to mirror the primary machine to the backup server
> - using EVMS to manage partitions and such
> - maybe using SSI to provide a single interface to the world for each 
> pair of virtual servers
> 
> Anybody have any experience with this kind of setup? Any suggestions, 
> pointers to recipes, etc.?

We run a similar setup here, but using VMware GSX/Server instead of Xen. 
  I've documented our (old, heartbeat v1 based install) setup at 
http://www.lancs.ac.uk/~clinch/dtvm/

We're now using heartbeat v2 for ease of rebalancing the cluster after a 
failure causes one physical host to take over the entire cluster, but 
I'm afraid I've not got round to documenting that properly yet.

We use LVM instead of md and EVMS in the following 'pile':

Virtual machine disk files
Ext3 FS
DRBD
LVM
Physical disk

Mentally I find it easier not to think of one server as 'primary' and 
the other as 'backup', rather both are 'compute nodes', otherwise you 
get funny looks from co-workers when discussing how your secondary 
server is now primary...

Hope that's some help,

Graham



More information about the drbd-user mailing list