RE: [nv-l] Netview SNMP/Discovery Issue

["Barr, Scott" <Scott_Barr@csgsystems.com>]

Make sure your solaris box has the init.snmpdx script coded with the "-f 0" parameter.

 

Basically, when NetView is using the communityNames.conf file, he probes snmp with "public" (actually with your "default" community name from xnmsnmpconf) and then each string on the list. The problem is that the solaris agent will literally terminate the snmp sub-agents when he receives 5 community strings that are not the word "public". This kills the sub agent including the one that  reports the interface table. This does not explain auto-deleting the object which I would assume is due to a duplicate interface address. Here is the init.snmpdx line I am referring to.

 

/usr/lib/snmp/snmpdx -y -c /etc/snmp/conf -f 0

 

I hope this helps. Just for the record, SNMP under solaris is um... a giant 2#%@#$@^@#$ pain.

 

-----Original Message-----
From: Lemire, Mark [mailto:mlemire@jhancock.com]
Sent: Tuesday, May 20, 2003 2:43 PM
To: 'nv-l@lists.tivoli.com'
Cc: Kenney, Jack; Hughes, Tom P.
Subject: [nv-l] Netview SNMP/Discovery Issue

We have encountered a situation where Netview is auto-deleting objects from its database almost immediately after discovering them.

This only appears to happen with Unix servers running Solaris 2.8; it's not a problem with our NT servers or with older versions of Solaris.

Here's our configuration:

(1) We're running Netview v6.0 on Solaris 2.6.
(2) Netmon is configured to use a seedfile that contains a statement: !@oid 0 to block auto-discovery of devices that are not running SNMP -- mainly workstations.

(3) Netview SNMP configuration does not have any specific or wildcard overrides in terms of this issue.  It uses a default community string that is configured on all routers and switches, which *do* get autodiscovered as expected.

(4) NT servers run an SNMP service but do not use the default community string.  These devices are not auto-discovered.  To manage these devices, we add their IP address to the seedfile once they are active on the network.  This works OK - once discovered, they stay managed.

(5)Unix servers running older versions of Solaris (i.e. 2.6) work the same way as NT; these servers do not run an SNMP daemon normally.

The problem is with Unix Servers running 2.8.  These do not use the default community string defined to Netview.  When we add the IP address to the seedfile, they do not get discovered.  When we explicitly add them in the read/write map, they initially get discovered but get auto-deleted 5 minutes later, apparently as part of the next polling cycle.  The only way to successfully get Netview to manage these servers (using the non-default community string), is to add a specific override in the Netview SNMP configuration.  We could also get around the problem by configuring netmon with the "-h" alternate community string option, but have rejected this solution since it would need to be left in place most of the time and adds to Netview's overhead.

We would expect this "auto-delete feature" of Netview to at least act in a consistent manner, affecting both NT and Unix servers alike.

Before opening a problem with support, we first just wanted to see if anyone out there has seen this or can offer an explanation.

Thanks!!

Mark Lemire
Sr. Consultant
John Hancock Financial Services