Filesystem full ?!?

[Benjamin Scott]

On Jul 3 at 12:07am, Drew Van Zandt wrote:
> fsck -f gives the filesystem a clean bill of health, with no change in 
> behaviour.

   Hmmmmm.  Curiouser and curiouser.

> Could it be disk geometry related?  I would think this would show up in 
> fdisk, though, not at such a late stage.

   Well, I'm not quite sure what you mean by "disk geometry".  If you mean a 
partition table that is somehow invalid (e.g., overlapping partitions or 
inconsistent entries), then I have found (much to my dismay) that "fdisk" does 
*not* always warn you of such things.  However, the kernel usually does. 
More significantly, fsck will notice (and usually quite loudly) if a bad 
partition table leads to a corrupt filesystem.  If fsck is quiet, one can 
usually safely assume the filesystem is internally consistent.

   You're sure you're using the right device file(s)?  I know this seems 
obvious, but a friend and I once spent about 15 minutes trying to figure out 
why "fdisk /dev/hda1" was complaining about a bad partition table...

   You're not using quotas, are you?

   Have you checked the kernel log ("dmesg") for any clues?

   What distribution and release?  What kernel version?  What version of 
e2fsprogs ("e2fsck -V")?  (Not that I know of any known problems with any of 
these off-hand, but the support tech in me won't let me not ask this stuff...)

   We should also back up a bit and make sure the error is really what we think 
it is.  Ideas in that vein:

   What command resulted in the error message?

   Try other commands.

   What was the full and exact text of the error message (cut-and-paste if you 
can)?

   Try other source files or destination directories.

   Run fsck on the source filesystem.

   I'm wondering if you're not trying to copy a corrupt file, or a device file 
that never ends, or if some temporary space somewhere is getting used up, or 
something like that.

-- 
Ben <dragonhawk at iname.com>