[GRLUG] lost+found

[Adam Tauno Williams]

On Sun, 2012-12-16 at 20:09 -0600, L. V. Lammert wrote:
> It looks like a large array we have has a number of lost inodes (ext4), as
> lost+found has gotten pretty big.

Have you seen any inode related messages in the log file?  [you should
configure your NMS to generate events on any such messages, if you
haven't already].

> Normally oI would just delete it and relase those inodes back to the free
> pool, but with "rm -rf", it complains directory not empty.

Is this commonplace?  Files in lost+found should be very rare.

Can you identify any of the files?  Are they sockets or 'files'?  If you
use "file" does it identify the content of any of the files?

If you have a growing lost+found then something is *wrong*.  Very likely
you have data somewhere that is now the bad kind of two buck chuck.

> What would be the correct way to clear lost+found? The partition is way
> too large to do an fsck, and it would be really nice to not have to
> unmount, if possible.

I'd consider performing an fsck to be absolutely mandatory.

-- 
Adam Tauno Williams <awilliam at whitemice.org>