[Net-snmp-patches] [ net-snmp-Patches-1926542 ] Trap: 'agenttrapaddr' config option to set v1 trap addr

[SourceForge.net <no...@so...>]

Patches item #1926542, was opened at 2008-03-26 17:14
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=312694&aid=1926542&group_id=12694

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Wayne F. Tackabury (waytac)
Assigned to: Nobody/Anonymous (nobody)
Summary: Trap: 'agenttrapaddr' config option to set v1 trap addr

Initial Comment:
Hi--this is a patch I've been meaning to submit for a LONG time--better part of a year, in fact.  I tried once, got kicked out of the system, and it skipped my mind until a good soul named Jan Safranek at RedHat poked, cajoled, and guilted me enough to submit this (truth be known, he submitted his own version of this as #1916840).

So, here's the original problem statement...

In 5.4, we were having trouble getting proper local addressing (agent-local) of outgoing coldstart and resource monitoring V1 traps.  

There are two issues here. 
One is simply getting the "clientaddr" directive (in the snmp.conf *for the agent's sent trap PDU*) to "take" on multihomed systems.  I think we made the same changes here that you made in patch #1775124.  However, if I get the gist of that patch, it will only affect the transport binding and, in essence, source address of the UDP packet itself.

There is also the matter of getting any agent address varbinds in the v1 trap PDU itself to be properly set.  I convinced myself that it would be overloading the intent of clientaddr to use the same parameter for this purpose.  I could see situations (e.g., firewall 
>> traversal) where one might want to set what's in the PDU payload, but not change the interface bindings or resulting UDP source addresses (or, indeed, not in general want to affect the clientaddr of real SNMP  client applications running on the same agent system just to address this issue).

To this end, I implemented my own snmpd.conf directive 
("agenttrapaddr"), which specifies a value to *only* set the in-PDU  varbind value for this.  If it's specified as a DNS name, I do the lookup at each point of trap generation, etc.

About this patch--I'm not sure what the "official" approved way to do this is.  Our source control isn't your source control.  So as you see it, I started with a 5.4 distro--last ChangeLog entry by Wes Hardaker (hi Wes!) dated 11/24/06 @ 09:44 if that helps at all.  I ran diff -Naur over a tree that was structured as 

net-snmp_5.4
      ./include.rel  /* The original 5.4 */
      ./include      /* My updates */  
      ./agent.rel    /* The original 5.4 */
      ./agent        /* My updates */

So, I ran two diff commands, resulting in the enclosed patch-friendly diff outputs.

We've been using this with great success for the better part of a year, as has (I presume) Mr. Safranek.

That being said, feel free to email me back at wayne(dot)tackabury(at)mirror-image.com if you have any questions or want to engage or abuse me (if that should start failing, try wayne(at)tiac.net).  If you want to chat or IM or something, email me and I'll give you a phone number to call.

Thanks!  Hope this provides something of worth to other people in the world!

Regards,
Wayne Tackabury
Mirror Image Internet


----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=312694&aid=1926542&group_id=12694