Free space wierdness

Nathan C. Burnett ncb at cs.wisc.edu
Thu Feb 5 12:55:10 PST 2004


I recently had a similar phenomenon happen on a Linux box, but the same
could happen on BSD I believe.

The problem ended up being that a process had created a bunch of huge
files, then deleted them but hadn't closed them yet.  The space doesn't
get reclaimed until the file is deleted and has no open file descriptors.
This is why the space came back after the reboot.  If it happens again you
can use 'lsof' (available in the ports collection) to find out what's
holding the descriptors open.

As for being over 100% capacity, I believe UFS (like most *nix
filesystems) reserves some amount of space that only root can use.  This
lets you boot and repair a system is an important filesystem (e.g. /,
/usr) is full.

Hope this sheds some light,
-Nate

On Thu, 5 Feb 2004, Herbert Wolverson wrote:

> I have a system running FreeBSD 4.9-RELEASE. It primarily functions as a firewall and
> router, and is generally pretty lightly loaded (load averages around 0.2). It
> is a low end system (P200, 64mb RAM, 2 gig hard drive), and is generally
> stable as a rock.
>
> The system has drives setup as follows:
> /	256M (UFS)
> /usr	1.2gb (UFS+Softupdates)
> (/var and /tmp are linked onto /usr/var and /usr/tmp respectively)
>
> This morning I noticed that the "/" partition was at 108% utilization,
> and "df -h" looked like this (approximately):
>
> Filesystem    Size   Used  Avail Capacity  Mounted on
> /dev/ad0s1a   252M   256M  -8M    108%     /
>
> Oddly, "du -h -d1 -x" showed only a total of 29Mb used on the partition!
> The output looked like this:
>
> su-2.05b# du -h -d1 -x
>  68K    ./dev
> 2.0K    ./usr
> 2.7M    ./stand
> 1.3M    ./etc
> 512B    ./proc
> 4.0M    ./bin
> 542K    ./boot
> 2.0K    ./mnt
> 6.4M    ./modules
>  30K    ./root
>  12M    ./sbin
> 4.0K    ./tmp
> 4.0K    ./oldvar
>  29M    .
>
> When I rebooted the system (without deleting any files), "df -h" showed
> the following:
>
> Filesystem    Size   Used  Avail Capacity  Mounted on
> /dev/ad0s1a   252M    29M   203M    12%    /
>
> This is good, since the correct amount of free space now shows, and the
> server is back to running perfectly. Can anyone shed any light as to why
> this discrepancy happened in the first place? I'd love to know what I can do to avoid ever having to worry
> about this again!
>
> Thanks,
> Herbert Wolverson,
> The Turner Stephenson Group, Inc.
> http://www.tsghelp.com/
> _______________________________________________
> freebsd-questions at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-questions
> To unsubscribe, send any mail to "freebsd-questions-unsubscribe at freebsd.org"
>


More information about the freebsd-questions mailing list