Actually this is math is "normal" to df.
The capacity number is correct, the used number is correct. But the available number is under-reported. It kind of makes you think your disk is full before it is really full. Note that even while non-root programs start to grapple with out-of-disk problems, programs running as root can still continue writing until the disk is truly full.
Regards,
.lzs
On Thu, 9 Dec 2004, Gene Heskett wrote:
Greetings;
I need a bit of help from the ext3 filesystem folks I think. I'm haveing a comprehension problem here. A snip from a df report:
/dev/hdd3 176100712 165509176 26340 100.3% /amandatapes
From the above, there is about 10GB on the missing list! Amanda justbailed out early because the disk was full, while the 2 figures on the left claimed there should have been somewhere between 10GB and 11GB of space left, not the 26 megs shown above.
The output of a du . on that partition, before I ran the amflush, was 165509176 bytes total used out of 176100712 that df claimed it has for a total, both in 1024 byte kilobytes.
So my question is: Where is the math broken? Both du and df, or my head?
Surely there cannot be 10+GB tied up in inodes and journals?
