I think the more common method is to have one file server with the rest of the machines saving files they want to share to the server. I'm trying to think of an advantage to your idea. Seems like it would adversely increase network overhead for machine A accessing a share on machine B, which in turn must access it from machine C. Better to have a central repository, possibly with partial or full mirrors on other machines. Then only one machine would need an NFS server configured, and the others could read from and write to that share as clients.
Another way would be separate specialized repositories on different machines, with clients mounting them directly to access shared files. Like machine A might serve all the music files, machine B documents, machine C something else. This might be better if you're worried about overloading a single server, especially if they are doubling as desktop boxes.
|