On Wed, Jan 10, 2007 at 03:06:05PM +0000, Dave Shield wrote:
>
> On 09/01/07, Lilach Givaty <[EMAIL PROTECTED]> wrote:
> > You are right, it isn't in the snmpd project. Sorry.
> > But maybe you can help me, when I do getbulk with max-repetitions very
> > large, I get a response from my device that is larger than a pdu (1500
> > bytes). The packet my device sends is fragmented.
> > From what I understood, the device should not send such a big packet, it
> > should send only the information that enters the first package.
> > Am I correct? Is this a bug, or are we misunderstanding the snmpV2
> > standard?
>
> I don't believe that SNMPv1 or SNMPv2c say anything much about the
> maximum size of a PDU. It's left up to the generating engine to decide
> how big is "too big". I think SNMPv3 addressed this concern, but I'm
> not sure.
There are two constraints involved here:
a) The agent should not send a response which the manager can't handle
because it is too big for the manager to process.
b) The agent may prefer to restrict responses to the PMTU to avoid
fragmentation.
Concerning a):
Only SNMPv3 provides the necessary information to an agent to
restrict response PDUs to sizes a manager can handle. In other
words, I believe a compliant SNMPv3 agent should not send response
messages larger than the msgMaxSize to the manager. (See RFC 3412
for a discussion of msgMaxSize and RFC 3416 for the protocol
operations and what to do if msgMaxSize is too small.)
With SNMPv1/SNMPv2c, this is not really specified well. But note
that snmpTargetAddrMMS (RFC 2576) may be configured to help out.
Most implementations support 1500 bytes these days, but if you
really want to be on the safe side, the only value you can really
bet on is 484 bytes (and I have seen engines being such
conservative).
Concerning b):
The SNMP specs are generally silent about this and I believe it is
really an implementation choice whether an engine does something
like PMTU discovery. In most local area networks, 1500 bytes still
is a rather safe bet - as long as you stay away from tunnels.
I recall that I once observed that the NET-SNMP stack does not really
honor the msgMaxSize field but this was quite some time ago so I
better do not comment on the current behavior.
/js
--
Juergen Schoenwaelder {International|Jacobs} University Bremen
<http://www.eecs.iu-bremen.de/> P.O. Box 750 561, 28725 Bremen, Germany
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Net-snmp-coders mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
