pkg/43166: net-snmp reports bogus CPURaw* values

[aaron.turner%equinix.com@localhost]

>Number:         43166
>Category:       pkg
>Synopsis:       net-snmp reports bogus CPURaw* values
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    pkg-manager
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 15 21:00:01 +0000 2010
>Originator:     Aaron Turner
>Release:        4.0
>Organization:
Equinix
>Environment:
NetBSD pulsar 4.0 NetBSD 4.0 (GENERIC.MP.TCPMD5) #2: Thu Jul 10 23:14:52 PDT 
2008  
cfiske%edrs-183.lab.equinix.com@localhost:/usr/src/sys/arch/i386/compile/GENERIC.MP.TCPMD5
 i386
>Description:
snmpd which is part of the net-snmpd 5.4.1 package starts reporting bogus 
CPURaw* statistics after a while (seems to be related to being queried 
repeatedly such as for system monitoring) which clearly aren't based on reality 
when compared to top or /kern/loadavg.  Specifically, CPURawUser approaches 90% 
utilization while CPURawIdle drops to 10%.  All other CPURaw* values become 
zero.

during the same period of time, /kern/loadavg reports values like:
18 53 0 2048

showing MUCH lower cpu utilization.

here's a report illustrating the problem 
DEBUG:  Values from first query:
DEBUG:  CPURawNice = 0
DEBUG:  CPURawSystem = 547
DEBUG:  CPURawUser = 7533754
DEBUG:  CPURawWait = 0
DEBUG:  CPURawKernel = 547
DEBUG:  CPURawIdle = 571396
DEBUG:  CPURawIntterupt = 0
DEBUG:  Total:          8106244
DEBUG:  ------------------------
DEBUG:  Values from second query:
DEBUG:  CPURawNice = 0
DEBUG:  CPURawSystem = 547
DEBUG:  CPURawUser = 7533888
DEBUG:  CPURawWait = 0
DEBUG:  CPURawKernel = 547
DEBUG:  CPURawIdle = 571404
DEBUG:  CPURawIntterupt = 0
DEBUG:  Total:          8106386
DEBUG:  ------------------------
DEBUG:  totals: 8106386 - 8106244 = 142
DEBUG:  delta CPURawNice: 0             0%
DEBUG:  delta CPURawSystem: 0           0%
DEBUG:  delta CPURawUser: 134           94.37%
DEBUG:  delta CPURawWait: 0             0%
DEBUG:  delta CPURawKernel: 0           0%
DEBUG:  delta CPURawIdle: 8             5.63%
DEBUG:  delta CPURawIntterupt: 0                0%

>How-To-Repeat:
query snmp oids:
%oids = (
     CPURawUser      => '1.3.6.1.4.1.2021.11.50.0',
     CPURawNice      => '1.3.6.1.4.1.2021.11.51.0',
     CPURawSystem    => '1.3.6.1.4.1.2021.11.52.0',
     CPURawIdle      => '1.3.6.1.4.1.2021.11.53.0',
     CPURawWait      => '1.3.6.1.4.1.2021.11.54.0',
     CPURawKernel    => '1.3.6.1.4.1.2021.11.55.0',
    CPURawIntterupt => '1.3.6.1.4.1.2021.11.56.0',
);
wait 60 sec, query again and do the math.

I have a perl script based on Net::SNMP which generates the above report.  I am 
willing to provide it upon request if it helps diagnose the problem.
>Fix:
N/A