Problem scenario
You run df and du. You see different amounts of CPU space that are free versus utilized. What should you do?
Possible Solution #1
Reboot the server. Rebooting may fix the problem.
Possible Solution #2
Run this command: lsof
Some files may be open that should not be. The lsof command should help you determine if a file's utilization is lingering.
Possible Solution #3
Do you know what directory is relevant? Through iterations of du and df commands, you can often determine what directory has the discrepancy.
Run this command (but substitute /foo/bar/ with the directory path in question):
fuser -v -m /foo/barPossible Solution #4
Can you unmount the file system in question (e.g., with umount)? If you can unmount it, run this command: Run this command (but substitute /foo/bar with the path to the file system): fsck /foo/bar
(This was adapted from https://www.ibm.com/support/pages/causes-mismatch-between-disk-usage-reported-df-and-du.)
Possible Solution #5 (less of a df issue but more of a du and quota discrepancy)
If you are having an issue with your disk quota settings and the du command results, see this Stackexchange.com posting.
Potential root cause #1
The root cause could be that an inode has been deallocated but some aspects of the system have not been updated (possibly because another process is using a file that lost its inode).
Potential root cause #2
A hidden directory could have a files with large sizes.
Potential root cause #3
An administrator with greater privileges could have isolated some directories with files that are taking up a good deal of space.