[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: df and du don't seem to agree ?



On Tue, Aug 21, 2012 at 04:07:54PM +0100, Karl E. Jorgensen wrote:
> Hi
> 
> On Tue, Aug 21, 2012 at 03:45:35PM +0100, David Cho-Lerat wrote:
> > Hi list,
> > 
> > this might be a newbie question, but can anyone tell me
> > why "du" and "df" don't seem to agree :
> > 
> > server:~# df -h /var
> > Filesystem            Size  Used Avail Use% Mounted on
> > /dev/mapper/vg00-var  5.0G  4.1G  624M  87% /var
> > server:~# du -h -s /var
> > 1.6G    /var
> > 
> > ("/var" is on a partition of its own.)
> 
> FYI: I always use the "-x" flag on du too, as this will not recurse down
> other mounted file systems - e.g. if you have /var/cache on a separate
> logical volume.
> 
> > "du" says 1.6G are used, while "df" reports 4.1G. Any idea why ?
> 
> "du" will only be reporting disk usage by files and directories it can
> find. It will not report overhead by the filesystem itself.
> 
> In your case, the discrepancy is too big (1.6G vs 4.1G) to be
> explained by this...
> 
> A couple of possibilities:
> 
> - Deleted files which are open: When a file is deleted, the space is
>   usually freed immediately. UNLESS the file is open, in which case
>   the space will not be freed until whatever-has-it-open closes it.
> 
> - Other mountpoints: If you have lots of files in a subdirectory
>   (e.g. /var/cache) and then mount another file system on /var/cache,
>   you cannot reach the original contents of /var/cache - and this will
>   then be invisible to "du"...
> 
> > 
> > I know some amount of space is supposed to be "reserved for the
> > super-user", but that's typically around 5%, right ?
> 
> Around that percentage, yes. But neither du nor df takes this into
> account.
> 
> > By the way, is there a command to see how big this reserved space
> > actually is on a given partition/disk ?

Correction: To find out the reserved-for-root space: I don't know. 

Unfortunately, I completely misread your question previously....


-- 
Karl E. Jorgensen


Reply to: