Re: Problem using net-snmp as subagent (Error in packet)

[Frank F. <fo...@ag...>]

Hi Dave,

Sorry, the below "issue" is a false alarm.
The user has encountered a firewall issue.
So NET-SNMP is properly using the authoritative
engine ID for discovery!

Cheers,
Frank

Frank Fock wrote:
> Hi Dave,
> 
> I have not forgotten to put together the AgentX
> issues that I have found while testing NET-SNMP
> interoperability. I think I can provide that
> report next weekend.
> 
> Meanwhile I came across another issue related
> to NET-SNMP:
> 
> SNMP4J(-Agent) supports engine ID discovery.
> A SNMP4J user reported, that he used a 5.3.1
> NET-SNMP tool to send an inform request to
> a SNMP4J based manager application.
> 
> The NET-SNMP tool sends a GET PDU
> to discover SNMP4J's engine ID.
> SNMP4J sends back a REPORT PDU with its
> engine ID in the authoritative engine ID
> field and the contextEngineID field is empty.
> NET-SNMP seems to ignore the authoritative
> engine ID and again sends a discover GET PDU.
> 
> If the contextEngineID is set to the
> authoritative engine ID then NET-SNMP
> seems to recognize the engine ID and sends the
> INFORM PDU. I am writing "seems to" because
> this has been tested with another tool which
> sends its REPORT from the same port as it
> received the GET PDU whereas SNMP4J is using
> a different port (and no contextEngineID).
> 
>  From my point of view this is a bug because
> RFC 3414 states in §3.2 that specifying the
> contextEngineID in a REPORT PDU is optional.
> 
> What do you think about this?
> 
> Cheers,
> Frank
> 
> Dave Shield wrote:
>>
>> We're currently in the (early) process of releasing 5.2.4,
>> so it's an ideal time to fix that particular line.  Fixes for
>> the 5.3.x and 5.4.x lines will take a little longer, but
>> I can get the patch applied so it's ready to roll out.
>>
>>
>>> BTW, I have implemented a few workarounds into
>>> SNMP4J-AgentX and AgentX++ to cope with some bugs
>>> of the NET-SNMP subagent. Are interested in the
>>> details?
>>
>> Most Definitely!
>> I've been pulling together the outstanding bugs on our
>> tracker system, so it would be good to get as much of
>> this sorted as possible.   A clear analysis of where things
>> have gone wrong is invaluable.
>>
>>>         There seem also to be 64bit OS related bugs
>>> when encoding an IpAddress on AgentX.
>>
>> That wouldn't surprise me either.
>> Let us have what you've got, and we'll see what we
>> can do.
>>
>> Dave
>>
> 

-- 
AGENT++
http://www.agentpp.com
http://www.mibexplorer.com
http://www.mibdesigner.com