[mrtg] Non-stop SNMP requests (was MRTG not updating RRD files)

Dean, Mike mdean at landstar.com
Thu Mar 8 15:33:22 CET 2007

Some additional information.  I've gone through the capture file and
this is what is happening:

At 07:50:49.615, MRTG starts its run with standard SNMP getRequests for
CPU and memory.  After that, it starts a series of get-next-requests:

1) get-next-requests/responses for the ifName tree.  A total of 332
request/response pairs
2) get-next-requests/responses for the ifDesc tree.  A total of 332
request/response pairs
3) get-next-requests/responses for the ifType tree.  A total of 332
request/response pairs
4) get-next-requests/responses for the ifAdentIfIndex tree.  A total of
41 request/response pairs
5) get-next-requests/responses for the ifPhysAddress tree.  A total of
332 request/response pairs
6) get-next-requests/responses for ifName, ifInOctets, ifName,
ifOutOctets.  For each interface, this request is made as a single
request.  A total of 283 request/response pairs are made for this set.

The last request/response pair for item 6 ends at 07:52:10.850.  At
07:52:10:857, the process is started over from the top.  From what I can
tell, this loop simply continues without end.  It never seems to be

Any ideas?


-----Original Message-----
From: mrtg-bounces at lists.oetiker.ch
[mailto:mrtg-bounces at lists.oetiker.ch] On Behalf Of Dean, Mike
Sent: Thursday, March 08, 2007 8:44 AM
To: mrtg at lists.oetiker.ch
Subject: [mrtg] Non-stop SNMP requests (was MRTG not updating RRD files)

I believe I've discovered why MRTG was not writing out the RRD files,
but answering that question, generated more questions and no real
solution as of yet.

I did a tcpdump/traffic sniff of the SNMP traffic going between my MRTG
system and one of the Cisco 6513 switches whose RRD file is not being

The SNMP get-requests and responses never stopped!  I capture the info
to a file and for 25 minutes (07:50:49 to 08:15:59) the SNMP traffic
literally did not stop.  The requests are getting responses, within .05
seconds but looking at the timestamps, there is a request/response pair
every .10 seconds and it did not stop for the entire capture period of
25 minutes!

This particular switch has 190 interfaces, but this type of traffic
should not be occurring.  The SNMP get-responses, BTW, are not timeouts,
they are response with valid data (numeric).

I did a traffic sniff of another 6513, with 103 interfaces that was
local to the MRTG box.  That SNMP traffic which included getting not
only the traffic stats, but the port error counters as well (double the
number of SNMP gets) took about 30 seconds to complete.  Note that the
switch with the problem is connected via a DS3 connection to another
site.  The DS3 averages about 10mbps of traffic so there is plenty of
room there and the average response time from the MRTG box to the switch
having an issue is 49ms.

So no the question is why the constant stream of SNMP traffic?


mrtg mailing list
mrtg at lists.oetiker.ch

More information about the mrtg mailing list