On 8/16/2011 12:30 PM, Camaleón wrote:
A single volume of 2 TiB is very big (and big file systems are more prone to errors and hard to recover in the event of a corruption... fsck can take... ages? :-P).
I've never seen a correlation between filesystem size and probability of FS corruption. FS corruptions are always due to singular events, such as hardware failures and kernel bugs, neither of which correlate to filesystem size. Thus, this is a non issue.
EXT3/4, XFS, JFS are all journaling filesystems. Thus the need to manually run a check will be very rare, especially in the case of XFS. Thus I wouldn't shy away from a single 2TB filesystem. The average XFS filesystem size of machines in the wild today is well over 20TB. There are many multi-hundred terabyte XFS filesystems in existence, and there exist CXFS filesystems over 1 petabyte.
Also, if you plan to host specifically multimedia files you could consider using XFS instead, I've been told is very good for such purpose ;-)
It is. But keep in mind that XFS is not a filesystem for noobies--not plug 'n chug. One need take some time to learn XFS. For instance, fsck.xfs doesn't exist. The proper command is "xfs_check" or "xfs_repair -n". This is but one example of why "learn then do" is the way to go, not the other way around.
-- Stan