Hi Everyone!

I'm looking to set up a very simple small office virtualised server system, where I have two physical servers, each hosting a number of virtual servers, and mirroring each others data.

In the minimum scenario, I want to be able to transfer the virtual servers from one host to another (preferably) automatically, and better still, load balance the servers from one host to another.

Almost every discussion of clustering etc. seems to require more servers, and/or be at a higher performance level than I am looking for currently. However, the point of virtualising and clustering all my server process would be that this becomes more easily scalable for the future.

If anyone has managed to do what I'm trying to do, and would share what technology and process they used, that would be really helpful.

I'm currently using Centos 5.5, but am perfectly willing to change distribution if there this is easier on another.

If this has been discussed a thousand times before, please forgive me and point me in the direction of any such threads.

Thanks.

Kheldar.

well terms like "clustering" can mean all sorts of different things. What scenarios do you want to enable? generally CentOS comes with clustering services, the rebrand of the Redhat clustering technologies, but you need to be clear what you mean by clustering. what kind of services etc.? generally what you have sounds absolutely fine, and probably simple solutions you can pick up on linux-ha.org might suit you fine. I've found this simple guide http://onlamp.com/pub/a/onlamp/2003/...inuxhacks.html of great use for a painfully simple failover model.

well, keeping the machines on two hosts would require a central storage. Once you have that in place, and accessible by both hosts, you can bring up RHCS with the libvirt-managed VMs as cluster services. So if a VM dies on one host, or one host goes down, the VM gets restarted on the other host.

well you could use drbd or something similar to not require a dedicated storage node.