You might want to have a look at DRBD (http://www.drbd.org). It's a
distributed, redundand block device which you can use instead of the
expensive iSCSI/AoE/FC shared storage backend to base your cluster FS
upon. It works quite well with good performance and excellent coverage of
possible error conditions.
Even without a cluster filesystem (and it's complexities), you can at
least build a good shared nothing failover cluster for your central