[mrtg] Wrong Values

Freddie Cash fjwcash at gmail.com
Fri Feb 27 19:40:37 CET 2009


On February 27, 2009 10:07 am mikelowrey wrote:
> I installed MRTG last week as described in MRTG 2.16.2 Linux/Unix
> Installation Guide.
> Everything works fine - I get data from my cisco fc switches and I can
> take a look of the statistics via browser.
>
> Today I compared the data from MRTG and the Fabric Manager Web client
> (from Cisco) and the differences of the collected data are pretty huge
> (up to 50%). In every case MRTG shows lower values.  The values are
> updated in both cases every 5 minutes.

Most SNMP counters are 32-bit, which means they have a max value of around 4 
billion (packets, bytes, bits, whatever it's measuring).  5 minutes is a 
*very* long time, and the counters can easily wrap around several times 
during the time period.  Which can lead to very misleading results.

Either shorten your polling period to around 60 seconds, or change your 
Target[] entries to use 64-bit counters (if they are available), or both.

-- 
Freddie
fjwcash at gmail.com



More information about the mrtg mailing list