I'm about to reorganize our raid array, and I\m thinking of building a big root filesystem containing essentially the usual root stuff plus /usr. The idea is to minimize the number of partitions to avoid seek on the disks. It's often said that the root filesystem should be as small as possible to make it less susceptible to errors. I know that with good modern scsi disks there's not much danger in having a big filesystem. What I'd like to know is about fsck. Will a big filesystem be more prone to fsck troubles if the machine stops uncleanly? This seems likely, but I'm interested in those files that are not often written to, only read, like in the traditional root arrangement. My hope is that the files that have already been synched are not vulnerable to fsck inconsistencies, only those that haven't been updated before the crash. Is this right?
