[lug] 'df' returning bad numbers

[D. Stimits]

Jeff Schroeder wrote:

> I just did a major housecleaning on my system, moving about 20GB of data
> to another hard drive.  Although I know the drive now has a ton of free
> space, when I use 'df' it reports a usage of 100%:
> ...
> /dev/hda6             69860376  68615112         0 100% /xdata
>
> /xdata is the disk in question-- you can see the data I moved to /data
> takes up 18.7GB.  I'm positive there's actually at least that much
> space free on /xdata.

...
I think it would depend on your inode count (thus inode density), and 
average file size. What is the average size of data files? What is your 
inode density? Use df with "-i" option for viewing inodes on both 
partitions, use "-h" if you want more human readable formats. What 
filesystem type is it (see the mkfs.whatever page for inode information)?

Example good scenario, inode density 4k/node, file average size 3.9k. 
Example bad scenario, average file size 512 bytes, inode density 
8k/node. Example bad scenario, average file size 5 MB, inode density 
4k/node.

D. Stimits, stimits AT attbi.com