ZFS and disk usage

[Freddie Cash]

On Fri, Apr 13, 2012 at 8:27 AM, Johannes Totz <johannes at jo-t.de> wrote:
> Without checking the numbers myself...
> Note that zpool and zfs do not agree on (free) space accounting: zpool
> shows "raw" space, whereas zfs includes metadata overhead for itself.
>
> Small rant: I dont understand why zpool and zfs show different things.
> If you have an integrated storage stack then why not show consistent
> numbers? Is there any use for this extra (mis-)information that
> zpool-vs-zfs provides?

There's a great posting about the differences in the zfs-discuss
mailing list archives, although I can't find a reference to it at the
moment.  Going from memory, the breakdown is something like:

zpool shows "raw storage available to the pool across all vdevs",
without counting any redundancy.  This should be approx. "size of
drives * num of drives".

zfs shows "storage space available for use", after removing all
redundancy, extra space for metadata, checksums, etc.  This is what's
available for programs to use, before compression and dedupe take
effect.

df shows "storage space available to userspace programs" after all
compressions, dedupe, metadata, checksums, etc have been removed.
This is the actual space that users can access.

"ls -l" shows the "size of files" (as in, uncompressed, rehydrated,
the size it would be if you copied it to a floppy).

They each work on different layers of the storage stack (DMU, ZPL,
userspace, etc).  Hence, they show different values.  But once you
think about what each layer of the stack is doing ... the numbers make
perfect sense.

-- 
Freddie Cash
fjwcash at gmail.com