server slows to a crawl, but 'ps' doesn't show anything

[Edward Vielmetti]

clio is a 3/260 running SunOS 3.5; it's the nd server for 8 systems
including bloch, a 3/50.

One morning clio is running very slowly; the red blinky light on the back
is going at about half speed, and vmstat shows roughly 50% cpu use even
when everything else is quiet.  Rebooting doesn't help.  Looking at the
output of 'ps auxw' doesn't show any unusual cpu hogs, which is quite
puzzling -- where is the processor going?

A look at 'nfsstat' shows a lot of nd traffic.  Take a peek at the net
with 'traffic' shows a lot of nd traffic, in particular between bloch and
clio.  Go upstairs to reboot bloch, and sure enough it's been dead since
about midnight with the message "proc: table is full".  Reboot bloch and
everything is back to OK.

Two questions:

1.  Why didn't 'ps' show some process being very active serving nd
    requests?  Which process(es) is (are) responsible for that?  (biod didn't
    have an unusual amount of cpu usage).

2.  What would cause an idle diskless machine to suddenly seize up
    with a "proc: table full" message?  No one was signed on at the
    time.

Puzzled,
Ed

Edward Vielmetti, U of Michigan math dept.  emv at math.lsa.umich.edu