Plugin cisco_qos uses 32bit counters instead if 64bit counters

Troubleshooting
, , ,
1

CMK version: 2.4 RAW
OS version: Redhat 9

Found out the hard way that the standard plugin cisco_qos uses the 32bit counters

using:
.1.3.6.1.4.1.9.9.166.1.15.1.1.9 (cbQosCMPostPolicyByte) and .1.3.6.1.4.1.9.9.166.1.15.1.1.16 (cbQosCMDropByte)

should be using:
.1.3.6.1.4.1.9.9.166.1.15.1.1.10 (cbQosCMPostPolicyByte64) and .1.3.6.1.4.1.9.9.166.1.15.1.1.17 (cbQosCMDropByte64)

On high bandwidth policies the 32bit counter wraps too soon and your graphs are screwed.

If you have this issue and want to fix now, edit the function snmp_section_cisco_qos:

Change this:

SNMPTree(
        base=".1.3.6.1.4.1.9.9.166.1.15.1.1",
        oids=[OIDEnd(), "9"],
    ),
SNMPTree(
        base=".1.3.6.1.4.1.9.9.166.1.15.1.1",
        oids=[OIDEnd(), "16"],
    ),

into this:

SNMPTree(
        base=".1.3.6.1.4.1.9.9.166.1.15.1.1",
        oids=[OIDEnd(), "10"],
    ),
SNMPTree(
        base=".1.3.6.1.4.1.9.9.166.1.15.1.1",
        oids=[OIDEnd(), "17"],
    ),

Also a good idea to delete the QoS* rrd and xml files of the Cisco devices using this plug in.

(as root) # omd su **
OMD[]:~$ omd stop nagios
OMD[]:~$ cd var/pnp4nagios/perfdata/**
OMD[]:~/var/pnp4nagios/perfdata/$ rm -f QoS_*
OMD[]:~/var/pnp4nagios/perfdata/$ cd
OMD[]:~$ omd start nagios

BTW this error exists at least since version 2.1 RAW