[Date Prev][Date Next] [Chronological] [Thread] [Top]


Ok, I didn't realise you were referring to AgentX support. I had some
bizzare image of building an SNMP daemon into OpenLDAP...

Kurt's point about an external process is valid - --without-threads is
(annoyingly) necessary in some circumstances. Perhaps you could use the
AF_UNIX support to do *LDAP* queries to the server, and get the information
that way? That has the advantage of leveraging much existing code. Perhaps a
RootDSE property set?


| Phil Mayers, Network Support     |
| Centre for Computing Services    |
| Imperial College                 |

-----Original Message-----
From: Lawrence Greenfield [mailto:leg+@andrew.cmu.edu]
Sent: 02 November 2000 20:45
To: 'openldap-devel@openldap.org'
Subject: RE: SNMP 

> Wouldn't it make more sense to have OpenLDAP provide the relevant
> information in a local format, so that an independent SNMP daemon could be
> used to aggregate all SNMP on the system (e.g. we use UCD and the SNMPv3
> security contexts requests, which we're hoping to use Kerberos auth on at
> some point...)

AgentX is an extensible, standardized SNMP framework.  The basic idea
is that the UCD master snmpd would answer the queries it knows about,
and the OpenLDAP subagent would register with the master agent.  Then,
whenever an ldap-specific SNMP query came in, it would be handed off
to the OpenLDAP subagent, which would return the requested data.

So adding SNMP (using AgentX) to OpenLDAP would complement the
existing SNMP on the machine, not replace it.

There are a couple of interesting questions about where the subagent
should run: inside the main process or as an auxillary process that
communicates via a shared file or some such.  I think implementing it
inside the main process would be somewhat easier and might be nicer
for some reasons (and not as nice for others), but would be a
reasonable start.

For the curious: here's the Netscape LDAP server documentation on it's
SNMP capabilities: