In my experience, MRTG "just works". Just recently, some strange behaviour has come to my attention.
I'ld appreciate any brief indication of whether what I'm seeing is a known problem, and if so, whether I can expect it to disappear following an installation of MRTG from current source. [ I have NO PLANS to attempt changes to the distro-supplied instance of Perl. 8-) ] The initial symptom was that my display front-end showed a number of units as unresponsive, on finding "NaN" in the RRD database. Investigation followed. Snmpwalk was able to return the counters which MRTG was not picking up. Tcpdump appeared to show that MRTG was not polling for some of the desired target OIDs. Moreover, tcpdump showed different traffic patterns between the same instance of MRTG and different probed units. In the "normal" traffic pattern, a succession of GetRequest PDUs are sent, each seeking four items of data from the probed unit (typically ifName.n, ifInOctets.n, inName.n, ifOutOctets.), for each interface corresponding to a Target in the configuration. For the units currently of interest, the count of associated 802.11 clients is sometimes specified in the configuration, and is duly indicated in the packets captured by netstat. The corresponding response PDUs also appear, as expected. When this "normal" traffic pattern is observed, the collected numeric data is stored in the corresponding RRD file, and the requested graphs are generated as expected. In the "abberrant" traffic pattern, some traffic is seen which matches the description given above. However, some expected GetRequest PDUs do not appear. In addition, a number of GetNextRequest PDUs are shown, stepping through ifName, ifDescr, ifType, ipAdEntIfIndex, ipAdEntNetMask, and ifPhysAddress. Again, corresponding responses appear. It seems that MRTG is trying to map the properties and status of the unit, as well as to collect data corresponding to (only!) some of the configured Targets. Data is stored as "NaN" for the configured targets which are omitted from the SNMP polling sequence. Corresponding graphs are, unsurprisingly, blank. This behaviour has been observed on Red Hat Enterprise Linux Server release 6.0, with the versions of Perl (5.10.1) and MRTG (2.16.2) provided in this distro. Our preference is to avoid (as far as is feasible) installation from source. The same MRTG configuration is also running without problem on an older system, due for retirement in the near future. This system runs Red Hat Enterprise Linux ES release 3.4, with Perl 5.8.0 and MRTG 2.15.2. Best regards and Happy Christmas! Niall O'Reilly _______________________________________________ mrtg mailing list [email protected] https://lists.oetiker.ch/cgi-bin/listinfo/mrtg
