High Load Averages Discovery.

Greg Patterson (root@gomer.mlink.net)
Tue, 7 May 1996 21:50:29 -0400 (EDT)

Some people have reported strange or abnormally high load averages (eg.
1.00+) for basically idle machines.

Tonight I noticed a similar situation where my load average was sitting a
little over 1.15. Usually it rests between 0.25 and 0.45 when not
compiling or processing something.

Looking at top or ps output didn't show any processes which could account
for the higher load average. And the CPU idle state still showed up as
basically 90% idle.

So, one by one, I closed applications (X, etc.) till I got back to the
console screens. Still nothing. Then I began to target various daemons
(httpd, kerneld, sendmail, inetd).

It turned out to be inetd but I don't recall it showing up frequently in
the 'top' listing.

So, was inetd actually using CPU resources, gone wacko, or is the load
average calculations getting messed up certain circumstances? This is all
with kernel 1.3.93 and an uptime of:

9:44pm up 16 days, 8:24, 8 users, load average: 0.39, 0.20, 0.28

I haven't gone past .93 since its been running smoothly since the day it
was compiled here. :)

Other information:

Slackware 3.0
procps 0.99a
libc 5.2.18
ld.so 1.7.14
gcc 2.7.2