> > I think you're going to need a larger hdd than that. To have an NFS-root > > diskless cluster, you really need separate copies of the OS storied on the > > master node for each of the slave nodes. > > You don't really mean this in its strongest sense, right? I mean, they might > need separate root and /var (can they safely share /var/lib and /var/spool?), > maybe a few different files in /etc (though most of that can be symlinked to a > common read-only /etc), but they can easily share /usr, which can account for > 60-80% of the disk usage, and of course /home.
We had some hard-to-track issues when we tried to only selectively share various directories, and it was much simpler to just copy the whole OS for each node. When hard drives are this dirt cheap, it's less of a hassle than tracking down the conflicts. Although technically you can break it up... just be prepared to have to fix wierd errors that are caused by files that don't like being shared amongst the clients... ~Warren
