🚀 go-pugleaf

 [ Our news server has been badly broken, so I'm resending this.
   Sorry if you've seen it before. ]
>>>>> "Gordie" == Gordie Freedman (gordie@duts.ccc.amdahl.com)

Gordie> Running df shows slightly different numbers, totalling around
Gordie> 360 meg.

Gordie> I remember on the Sun that you could set up a partition so
Gordie> that 110% of it could be accessed. 100% meant somewhere around
Gordie> 90%, so that root could handle problems with file system full.

df shows 4 sets of numbers, i.e.:

  Filesystem         1024-blocks  used available capacity Mounted on
  /dev/sd0a              14707    5304     7932     40%   /
  /dev/sd0f             141625  103889    23573     82%   /usr
  /dev/sd0d              14707    2447    10789     18%   /export/root
  /dev/sd0e              77567   71737    -1927    103%   /export/swap
  /dev/sd0g             280479   64243   188188     25%   /priv/orange

First we get the device (or remote) name. The comes the total number of
blocks (i.e., total capacity of partition), space used on partition,
space left on partition, percentage of disk used, and finally mount
point.

The first number *is* the total size. If you add used and available,
you'll see that it does not add up to the total size. Instead, they
add up to 0.9*total. The remaining 10% is "reserved". The "percentage
used" is of the 90%, i.e., the 40% in the first line above means "40%
of the space available when the partition was created has been used"
(this is really 36% of the total partition).

Now for the "reserved" 10%. This space is use to make the Berzerkeley
Fast File System "fast", that is, to have space to group blocks
reasonably, to avoid fragmentation, etc. In most situations, using
this space will make disk accesses slow -- if you draw a curve of disk
speed relative to percentage used with FFS, you will get a nearly flat
curve out to approx. 90%, where the curve rises sharply.

In an ordinary partition, only the superuser can use this space, and
doing so should only be done in special situations, e.g., when
creating one big file in one go (linear file) and so on. If this is
done, df will list the partition as >100% full (as for /export/swap
above).

So, with FFS we loose 10% of the disk space. This is the price we pay
for the good performance of FFS and the freedom from running
defragmentors and other silly stuff. You can format a partition to
avoid this (the minfree option to mkfs), but the hit you take is big
and the practice is not recommended.

/Lars
--
Lars Fischer, fischer@iesd.auc.dk | It takes an uncommon mind to think of
CS Dept., Aalborg Univ., DENMARK. | these things.  -- Calvin