Had a issue recently where DB admins did:

mv listener.log listener.log.old
gzip listener.log.old

Where file listener.log is continually being written to by an oracle process.

Anyway, according to the DB admins they successfully created the
listener.log.old.gz
file (i.e. all went okay).

Then they started complaining that space had not been freed on the filesystem.

I logged in and couldn't see the file listener.log.old.gz anywhere. The "listener.log", which was 2GB originally, was indeed smaller when I checked, so it was moved.

"du -ks /opt/sw/oracle" showed 4GB in use.

Whereas "df -k" showed 6GB in use on the "/opt/sw/oracle" mountpoint.

Apparently it's something to do with inodes referring to the original file while it's still in use ... but not sure, which is why I am posting to find out why?

Anyway, to fix the issue, stopping and restarting the process using the file will free up the space.