disk_free_space and df, huge difference in output

Question

We have an application which stores files and provides a "disk filling level" to the user.

To achieve that we use the disk_total_space and disk_free_space combo, works great on the dev machine (mac os, local hdd around 250Gb).

On the production machine which use a 20Tb SAN bay storage the results are (way) off :

df -h output :

Filesystem                 Size  Used Avail Use% Mounted on
/dev/mapper/VG1-lv_foo01   20T   141M 19T   1%   /foo01

df -hi output (inodes) :

Filesystem                 Size  Used Avail Use% Mounted on
/dev/mapper/VG1-lv_foo01   320M  30   320M  1%   /foo01

disk_total_space("/foo01") returns 21902586179584 which is roughly 19.92Tb, looks about right.

But disk_free_space("/foo01") returns 20802911117312 which is roughly 18.92Tb, that would mean about 1Tb is being used on disk but in reality only about 140Mb is really used !

From where could that difference come from ? Inodes should not amount for a lot of space since only a few are used ...

I'm pretty confused, any idea ?

Answer 1

EXT systems reserve some space for root, by default I believe it is 5%. That must be why you see about 1TB used.

You can reduce that share or disable it with tune2fs:

#Set it to 1%
tune2fs -m 1 /dev/mapper/VG1-lv_foo01