You should also consider doing userspace NFS4 on top of DRBD partitions as your cluster server. It does not scale as well as some other solutions, but it may have other advantages that you want (maybe better locking, maybe better fail over support...). In particular, I would check out Jeff Garzik's server:
http://linux.yyz.us/projects/nfsv4.html
and the Ganesha server:
http://nfs-ganesha.sourceforge.net/
For some reason these servers seem to be relatively unknown but I believe they have a lot of potential...
Of course, I am holding out for ceph, it is the one solution that I consider 'correct', if only it could handle a 2 node server system properly (without being less reliable than a 1 node system)!
-Martin
Received on Thu Jul 29 15:35:49 2010