Rodolfo J. Paiz wrote:
At 14:44 4/22/2004, you wrote:
Root normally reserves space on a partition to prevent the disk filling up totally and crashing the system, but that's normally just 5%, so where's the other 5% (100GB) going? You can check how much space is "reserved" for root with:
With this in mind, how do you figure such a large space consumption for overhead? I interpret that as minimal and normal overhead -- formating factors, etc.
In fact, on my 30G physical drive as a single partition it reports a filesystem size of 27.94G (a loss of ~7% due to the differences in the way it is stated + overhead)
You'll note that that 1000/1024 is just about 2% short. That should be all you lose to disk naming conventions, but you may lose other space to (as others mentioned) inodes, overhead, etc. So you get to 27.94GB *filesystem size*. But if you add the "used" and "available" numbers from "df -m", do they add to that? Mine don't... they add up to about 5% less than the filesystem size. That's the reserved blocks.
On one disk I could check very quickly, I have this (edited for brevity):
<snipped>
AFAICT you are correct as relates to the management and overhead of the filesystem.
root@rita [~]# df -m Filesystem 1M-blocks Used Available Use% Mounted on /dev/hda2 55236 29532 22898 57% / /dev/hda1 99 7 86 8% /boot none 243 0 243 0% /dev/shm
So, the loss of space to which you refer is independent of, and additional to, the one I mentioned. In reality,
* Filesystem size = used + available + reserved
* Disk space = filesystem size + overhead
We are both right. <grin>
However, my point was not directly related to the overhead.
The OP on this thread (IIRC) said he had created a raid5 array from 9 250G drives, and expected to have a 2T array for use.
As I said earlier, the mfgr says the drive is 250G, but that is 250 X 1,000,000,000 (the mfgr definition of 1G). In terms used by most mfgrs 1G is actually ~931M due to the differences in the stated size and the size in binary terms. Thus each drive is a maximum of ~233G. 8X233 = ~1.864T which is then further reduced by the overhead of formatting, inodes, reserved, etc. as you have stated.
Yes, we are both right. 5% lost due to differences in stated size vs binary size, + 5% lost due to overhead of various types, easily makes this 2T array into only 1.8T formated and usable.
I was pointing out that a mfgr stated size is usually not the actual usable size of the drive, and you addressed the actual math to show how filesystems use space as overhead.
There are a few cases where the mfgr states his drive is XXG formatted, but that in my experience is rare, and would also usually be based on a dos/windows format and not *nix formatting.
On the box for an 80G Maxtor drive I recently bought it states "A gigabyte (GB) means 1 billion bytes" . Their stated definition of 1 gigbyte is 1,000,000,000 bytes, but gigabyte in binary terms is 1,073,741,824 bytes. The actual size of this drive as seen by the operating system is then 74.5G or less.
