Request/responses with proxy forwarder

[Ram Krishnaswamy]

Hi Jochen,

Here is the graphical representation.

(responses) dialup <-- Back-channel <-- Agent <--- Satellite (requests)
                              |
                              |
                         Dell, NAS agents

The request comes in through the satellite, the agent satisfies the request
if it has the data or forwards it to Dell or NAS proxy if it is intended for
those agents. The Dell agent is within the same box but the NAS agent is
outside the box but on the same LAN as the main agent. Both these agents
support SNMPv1. The problem is I need to differentiate each request to see
whether it is intended for a proxy agent versus the main one (Agent in the
above figure). Pay attention to the arrows too. Except for the communication
between agent and proxy agents all others are uni-directional. 

If there are no proxy agents, I am able to get the requests and send the
responses through the back-channel without any problem. But when the request
is intended for a proxy agent, I need to be able to differentiate and not
send it to back-channel manager but instead directly send it to the proxy
agent, get the response and then forward it through the back-channel
manager. 

Do you think your second solution, extending pdu/target/address class, solve
this problem? If it does, can you describe it in a little bit more detail? 

Thank you very much. 

Ram                              


-----Original Message-----
From: Jochen Katz
To: Ram Krishnaswamy
Cc: 'agentpp-dl____agentpp.com'
Sent: 12/14/2001 7:15 PM
Subject: Re: Request/responses with proxy forwarder

Hi,

actually I was unable to understand the complete scenario from the
textual description. Maybe a small ASCII graphic with node names could
help.

> One solution that I was thinking of was to let SNMP++ code know that
the
> request is for a proxy agent (thereby bypassing back-channel manager).
But
> the question I have is what happens to the response from the proxy
agent?
> How do I differentiate between proxy agent request/response versus
sending
> the response to back-channel manager? If this is possible, how can
this be
> done? Any other ideas would be greatly appreciated.

The first thing I see is a list of source IP-addresses that could be
used to decide if the response should be sent back or forwarded to the
back-channel. A different solution could be an extension of the
Pdu/Target/Address class. An additional flag indicates that the
back-channel should be bypassed. You would then have to modify agent++
to set this flag for proxied requests.

Kind regards,
  Jochen