Thanks for the reply.  The Simple_Async_Application in your link uses one
session/socket/SRC port per agent.  I've compiled and run this sample
successfully, but I haven't been able to modify it to use a single
session/socket/SRC port.

I understand "don't hammer agents" but I don't understand your "one query
per agent" limit - is this a limitation of the API?

I'm new to this API, I might be missing key concepts... but I am confused
by the "traditional vs single" distinction, and I'm curious what
"subsessions" are.







On Tue, Jan 5, 2021, 4:26 PM Craig Small <csm...@dropbear.xyz> wrote:

> Hi,
>   Wouldn't asynchronous queries do what you need? See
> http://net-snmp.sourceforge.net/wiki/index.php/TUT:Simple_Async_Application
>
> I've written something similar but was using pysnmp and it seemed to do
> the job.  The trick is you also have to make sure you don't hammer the
> agent; most agents are pretty awful and lock up if you hammer them too
> much.  For the reuse of the UDP port, you might need to go to socket level
> ioctls and use SO_REUSEPORT. The danger is you can have only one and only
> one query per agent.
>
>  - Craig
>
>
> On Sat, 2 Jan 2021 at 12:32, Ed Fair <quacksp...@gmail.com> wrote:
>
>> I'm trying to write an SNMP manager using the net-snmp API.  The manager
>> will use UDP in IPv4 and IPv6, and will be sending get-requests (and
>> getbulk, and getnext) to 10000 agents (arbitrary number, but you get the
>> idea - a relatively large number).  Is there any way to do this using just
>> a single session?
>>
>> I've compiled and experimented with the demo apps, they seem to use one
>> session per agent.
>>
>> Creating 10000 separate sessions seems excessive (as I understand, this
>> would consume 10000 sockets and 10000 distinct SRC PORTS from my host).
>>
>> Creating/using/closing 10000 sessions seems excessive as well.
>>
>> The reason I'm going here:  my manager must interact with a firewall
>> before sending any requests to any agents; a single firewall interaction
>> identifying a single UDP source port would be much more desirable than
>> 10000 firewall interactions identifying 10000 different UDP source ports.
>>
>> Thanks in advance for your comments.
>> _______________________________________________
>> Net-snmp-coders mailing list
>> Net-snmp-coders@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
>>
>
_______________________________________________
Net-snmp-coders mailing list
Net-snmp-coders@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/net-snmp-coders

Reply via email to