Not a comprehensive answer, but some clues, maybe....
You will end up with an array of interfaces on the gateway router that you
cannot ping,
and the matching set of red subnets. There will be a finite number of
these interfaces,
and if you delete them they will come back. All I have been able to figure
out to to
with them is to unmanage and hide them. Usually I can define a collection
of those
interfaces to remind the customer which ones to keep unmanaged and hidden.
Putting addresses that you cannot ping in the seedfile does not cause them
to be
discovered. Netview will only add the addresses it finds in the address
table for
the gateway router. That will include the address(es) on the local side of
the router.
The local address is the one you would put in the seedfile.
No snmpget is done during discovery until AFTER Netview is able to ping an
address. Therefore no attempt will be made on the unregistered addresses
because
Netview cannot ping them. Even if they are in the seedfile.
Netview will discovery and create nodes for any unique adddresses it finds
on
devices with duplicates, and then complain about the duplicates. But I
don't think
you will get into that situation with the unregistered addresses.
Cordially,
Leslie A. Clark
IBM Global Services - Systems Mgmt & Networking
Hello,
First let me thank you for your numerous answers about map copying. We'll
need
to check with this customer how he can change his whole procedure.
I have found this post in the internal IBM Netview forum. As it still
didn't get
any answers or comments and I think this question is quite interesting,
please
let me ask it on behalf of the original sender.
Hi,
I am investigating the possibility of using Network Address Translation
to allow management of multiple networks with overlapping (unregistered)
IP address space. I am curious how the NV discovery process will behave
when using NAT to map the addresses to registered ones (1-to-1, no pool).
What will happen if we put a registered address in the seedfile and let
NV discover the system? The answer to the SNMP get request during the
discovery will only contain interfaces with unregistered addresses, and
will NOT contain the registered address it sent the SNMP request to, so
will NV still add an interface with the registered address to the node?
Because of the overlap in the unregistered address space of the networks
NV will discover (through the SNMP get request) duplicate unregistered
addresses (although it cannot ping them). Earlier experiences with
NV4 and NV5.0 have made clear that this means that the entire node
with the duplicate address (not just the offending interface) is not
added to the object database, or, when the node does get added, it leads
to database corruption. Has this been improved in NV5.1?
Does anybody have any experience with these issues?
Best regards,
Frederic Mottiat
|