On Mon, Mar 29, 2010 at 10:13:54PM +0200, Oskar Svadkovskey wrote:
I was referred to you from Wireshark. Wireshark does not parse our V2 traps. Recently I noticed that it does not parse Veritas Cluster traps either.
The error I see is "A scalar should have one instance sub-id this one has none"
Indeed if the varbind is suffixed with .0 or .whatever, then wireshark can parse the varbind correctly.
I see that traps sent manually with snmptrap do have SNMPv2-MIB::sysUpTime.0 and SNMPv2-MIB::snmpTrapOID.0
If I add this .0 to our varbinds manually, everything is fine
My question is if this failure to parse indeed comes from libsmi. I should notice that everything worked perfect with Ethereal that uses the previous version of libsmi.
Simply google for the error message and you will find where it comes from:
http://anonsvn.wireshark.org/wireshark/trunk/asn1/snmp/packet-snmp-template....
So this is clearly a wireshark error message, not a libsmi error message.
I am not sure I understand your description correctly but it sounds like the trap contains a varbind OID which matches exactly on OBJECT-TYPE definition. If my guess is correct then the agent sending the trap is indeed not behaving correctly (and it does not matter whether other agents show the same erroneous behaviour).
/js