[RADIATOR] SNMPAgent and high peak in cacti graph at logrotate

Dennis Ortsen dortsen at gmail.com
Thu Oct 2 03:39:15 CDT 2008


Hello all,

we have 2 radiator 4.3.1-1 (RPM) instances running on 2 RHEL5.2 server. 
On both we have enabled the SNMPAgent to monitor Challenges, Requests, 
Accepts in Authentication and we monitor Authentication errors in cacti 
(rrdtool). The SNMPAgent works fine, monitoring in cacti also works 
fine, but every night at 04:00 all graphs in cacti show a huge peak. 
This is exactly the same time that the servers rotate their logfiles. 
When radiator gets its logfile rotated, a SIGHUP signal is sent to the 
radiuspid to force radiator to use the new logfile. My guess is that 
this SIGHUP signal is causing the SNMPAgent to "reset" itself too, which 
results in the huge peak in the cacti graphs. This behaviour renders the 
cacti graphs unusable, while it is actually a nice collection of 
statistics we can use to better understand authentication errors in our 
wireless network.

Is there a way to get rid of that peak, in other words, make sure that 
the SNMPAgent doesn't get "reset" or skips the result of the peak?

SNMPAgent config section:
==========================================
<SNMPAgent>
         ROCommunity some_string
</SNMPAgent>
==========================================

current logrotate script for radiator:
==========================================
/var/log/radius/logfile {
     rotate 2
     size 50M
     daily
     postrotate
         /bin/kill -HUP `cat /var/log/radius/radiusd.pid 2> /dev/null` 
2> /dev/null || true
     endscript
}
==========================================

Thanks in advance,

with kind regards,

Dennis Ortsen
HAN University






More information about the radiator mailing list