I see something new at every single site I go to. Here's this week's:
The DNS had default name resolution set up for each subnet.
This apparently was to prevent unresolved names from ending up
out on the internet or something; a perfectly valid reason, so we had to
live with it. Of course many of the network devices were not in DNS, so
the addresses all resolved to a handful of names like 'subnet-xx.xx'.
And there would be no forward lookup for this name, obviously.
The result was that Netview made an object for every one of these
default names, most of them with lots and lots of interfaces. And, it was
spending a lot of time deleting and adding these nodes and interfaces,
probably because of the lack of forward name resolution
To work around this, first we limited discovery by negative address
ranges as strictly as possible. Then we excluded non-snmp nodes by oid.
Then, for each object that was still created with that name, we had to
investigate every address on them with snmpwalk, figure out which
addresses went on which device, and add entries to /etc/hosts file,
then delete and rediscover them. The /etc/netsvc.conf specifies
'hosts=local,bind' so the hosts file is checked first.
This will be an ongoing process for this customer until all devices have
been named. A Smartset to detect the arrival of new ones would be
in order. You know, it might have been easier to extract the named
nodes, build a hosts file, and disengage from the DNS. The next time
I run across this, I think I would take that approach first.
Cordially,
Leslie A. Clark
IBM Global Services - Systems Mgmt & Networking
|