I have a load balancer and several servers behind it. I'm looking for a multi master file replication tool that would allow me to write on any of the machines and replicate the content on all the others, on an event-basis, in a very small timeframe (5 sec).
A single NFS volume mounted on all servers will not suffice, because it becomes a single point of failure.
I'm running CentOS on the servers, the load balancer has proprietary software.
I use DRDB with OCFS2 as my filesystem.. works great.
The mistake a lot of people make with drdb is they use a non-clustered filesystem. Oracle packages ocfs2 for redhat/centos for all the kernels for v5
Have you thought about block-level replication with DRDB?