[AGENT++] timeout

[Don Radick]

Not much we can do, but a lot you can (and should) do.

SNMP is tcp/ip protocol UDP.  That means nothing intrinsic to
the network stacks (on the client and server) will make any
attempt to track open requests, and how long since the response 
has been sent.  And since the snmp daemon is under control of the
cpu of the remote host, if it's real busy, you may not receive 
any reply.

But you can do this in your own network code, or maybe find 
somone else's "tcp over udp" library you can use with SNMP++

Sounds like you have a "busted" snmp daemon.  Get-Bulk has been
standard for many years.  It _greatly_ increases efficiency if you
are retreiving lots of information.

regards,
-Don



On Mon, 2006-08-28 at 13:55 +0200, Manuel Werlberger wrote:
> Hi!
> 
> I'm working on a monitoring software for many snmp variables. In fact 
> snmp++ works gread polling all the snmp variables with the asnychronous 
> get methos. But one thing brings some real troubles. If we have many of 
> snmp-vars to poll (which is in fact all the time) i get timeouts from 
> some vars. In the next polling round they jump back to their desired 
> values. Rising the timeout value doesn't bring any advancement.
> 
> The bad thing is that every time a timeout occurs there should be an 
> alarm which could maybe be a little bit annouying for the people using 
> the software afterwards ;-).
> 
> Is this a known issue? Does somebody have some advice or tip for me?
> 
> One thing to say that unfortunatly our snmp-deamon cannot handle bulk 
> requests.
> 
> Rgds,
>  Manuel
> _______________________________________________
> AGENTPP mailing list
> AGENTPP at agentpp.org
> http://lists.agentpp.org/mailman/listinfo/agentpp