[AGENT++] Multiple MIB instances

Andreas Vinsander andreas.vinsander at teligent.se
Thu Apr 22 14:16:20 CEST 2004


Marek Malowidzki wrote:

>----- Original Message -----
>From: "alexander link " <link.alexander at firemail.de>
>To: <agentpp at agentpp.org>
>Sent: Thursday, April 22, 2004 10:47 AM
>Subject: [AGENT++] Multiple MIB instances
>
>
>  
>
>>Hi.
>>For our enterprise it is needed to manage our own software servers with
>>    
>>
>snmp. Due to security reasons we'd
>  
>
>>like to use version 3 of this protocol. We also want to utilize agent++
>>    
>>
>API because it seems to work very good
>  
>
>>and stable also.
>>It may happen that more than one software servers are running on a single
>>    
>>
>machine with ONE IP address.
>  
>
>>Thus an implementation is needed to handle this issue.
>>So is it possible to manage all of these same servers with agent++?
>>It is not known how many and when these servers come to life. This depends
>>    
>>
>on customers need.
>  
>
>>So is this problem in general solvable with your agent++ API?
>>In this context, can someone give a short instruction guide on things to
>>    
>>
>pay attention to?
>
>If I understand correctly, you need to design a MIB for your servers and
>build an SNMP agent to collect info about the servers (and perhaps even
>allow to perform some operations on the servers, although SNMP is not a best
>solution here). Agent++ (and any other SNMP agent implementation) does not
>provide any particular solution for the problem - as I can guess - you need
>to do it yourself. For example, a server could notify the agent about its
>presence using some custom protocol (e.g., sockets, shared memory,
>anything), perhaps periodially refresh its state (or allow agent to poll for
>it). If fact, the custom protcol could be SNMP as well (e.g. servers are
>allowed to update data while other clients are only able to read state),
>especially with SNMPv3.
>
>Servers could be kept in some table (one row per server), that would group
>all the common data. Additionally, to keep specific data (specific for a
>given server type), augmented tables could be used. Looks complex but is
>indeed simple. This is first that comes to my mind.
>
>Only one IP address or more addresses does not matter at all.
>
>Marek
>
>  
>
Why not make all the server agentx subagents? Each of them responsible 
for its own OID range.

/Andreas




More information about the AGENTPP mailing list