On Mon, Nov 01, 2004 at 02:35:42PM -0500, Thomas Ledbetter wrote:
> Thanks for the reply Nathan - I am really confused here..
> I have a 137G production xfs filesystem, of which 79% of the blocks are
> and 11% of the inodes are used. The filesystem used 4K blocks.
> We seem to be on the verge of some hidden capacity limit as we keep getting
There aren't any hidden capacity limits - either the space is in
use by something (like open, unlinked files as someone suggsted),
or some free space accounting has gone awry inside XFS.
> 'no space left on device' even though there is 21% (30G) of free space.
> What could this possibly be? Its a RedHat 9 system with the kernel upgraded
> version 2.4.25, so its not cutting edge 2.4.. but its not all that old
Did you try the unmount and xfs_repair I suggested earlier?
What did xfs_repair have to report?
> We are using xfs extended attributes heavily.
Out of curiousity, what would the average attribute name and value
lengths be? And what does your xfs_info report look like for this
> From: "Nathan Scott" [nathans@xxxxxxx]
> Date: 10/29/2004 03:43
> On Thu, Oct 21, 2004 at 12:30:37PM -0400, Thomas Ledbetter wrote:
> > Does anyone have any ideas why this would occur or where I can turn to for
> > more information?
> Unmount and run repair - does that show any disconnected inodes?
> (if so, they'd be put in lost+found - these may account for your
> missing space, if any exist..).