[GLLUG] NFS is worn out?

[John Winters]

Can anyone give me a little guidance on a rather odd problem.

I have a SheevaPlug running Debian Squeeze, which has been happily 
running as a domestic file server for longer than I care to remember. 
Just recently it's developed a bit of a problem, but I can't see 
anything to account for it.

The two symptoms are:

1) System load (as shown by top) rises, first to a steady 1.00 and then 
in time to a steady 2.0.  However the CPU continues to show ninety-odd 
percent idle time and no task appears to be using a lot of CPU.  Even 
when I use htop and ask to see kernel threads, nothing accounts for the 
extra load.  The only way I can get it back down is to re-boot the box, 
after which it will revert to a load around 0.00 for a while (possibly 
several days).

2) NFS breaks regularly.  May be associated with the rising system load, 
but I haven't managed to pin this down.  The symptoms are that it gets 
very slow at serving requests, particularly those involving file 
locking.  Opening a file on the server using LibreOffice on a client is 
a good way of getting the problem.  The only problem logged is:

May 31 17:57:50 sheeva kernel: [184235.870895] statd: server rpc.statd 
not responding, timed out
May 31 17:57:50 sheeva kernel: [184235.876782] lockd: cannot monitor revo2

"revo2" is the name of the NFS client trying to open the file. 
Re-starting NFS ("/etc/init.d/nfs-kernel-server restart") doesn't 
complete.  The only way to get it going again is to re-boot the box.

This box has been giving rock-solid service for years, and I don't see 
any hardware issues being logged.  The kernel was upgraded recently, 
along with a number of libraries, as part of the latest Squeeze security 
updates.  Apart from that as far as I'm aware nothing has changed.

The two symptoms may or may not be related.  Can anyone make any 
suggestions on how to investigate further?

TIA,
John