Hi,

I've got two NetBSD boxes (1.5.2 and 1.5.3_ALPHA) that have to deal
with fair amount of NFS traffic both as server and as client.  Box
boxes are Athlon 1.33GHz/512M RAM/60G HDD/ex&fxp NICs. Both boxes are
connected to 100BaseT switch (there is no packet loss at all).

The problem is that under heavy NFS load (mostly as NFS client with
NetAPP filer as a server) the boxes lock up for up to several
minutes. The boxes are pingable from outside, but that's about the
only sign of life. Increasing vfs.nfs.iothreads to 16 and running nfsd
with "-n16" didn't help much. When the boxes unlock, there are usually
a lot of runnable processes in the queue and load jumps up to around
50. It appears that something happens inside the kernel that
completely stops process scheduling.

Did anyone experience similar issues with NetBSD in NFS environment
and if so, what did you do to solve them?

Any suggestions on optimizing NetBSD box for better NFS performance?

Thanks,
--Artem