On Mon, 2005-05-23 at 10:50 -0500, Miles O'Neal wrote:
> About a week ago we brought an SL302-based
> file server online with a 3ware SATA RAID
> controller.  Since then we have had several
> of our NFS clients (compute servers) lock up
> after a kernel panic with a bunch of these
> errors on the console:
> 
> VFS: busy inodes after unmount: self-destruct in 5 seconds.  Have a nice day...
> 
A Google search turns up lots of this...

As an aside, not on Scientific Linux but on a RH 7.3 I have recently
spent a long time doing performance measures on a 3Ware SATA controller.
It does not perform well when there are many simultaneous writes going
on. This is a problem seen by many people - the I/O wait times go very
high. Not cured by latest 2.4 series kernel/elevator tuning etc.
Clients often report NFS timeouts in the logs.

Can I ask if you are mainly reading or writing data to the RAID
controller? Use iostat to see what is going on.