On Fri, 6 Dec 1996 22:40:07 +0100 (MET),
Simon Leinen <simon@switch.ch> said:
Simon> [...] When I ran the discovery tool on our backbone network,
Simon> it generated one node for each router *interface* (of which
Simon> there are hundreds )-:
Simon> [...] Can I do anything to help the discovery process find
Simon> out which interfaces belong to the same router?
Well, a look at the ip_discover.tcl script leads to the impression,
that the discover tool tries to do its very best (but maybe it should
do it like Avis: We try harder ;-)
Anyway, the question seems to be: How can two interfaces with
different ip-adresses on the same host identified to belong to this
node ?
The best practice would be a clear mapping of one hostname to multiple
ip-addresses along with the reverse mapping of each ip-address to this
one hostname. This needs to be managed via DNS and should help the
discoverer a lot.
However, a side effect would the the appearance of all ip-addresses in
the DNS, not honoring the always up-interfaces versus somtimes-up
interfaces leading to the known problems when using DNS answers given
in round-robin fashion (or problems with a named running on such a
router).
An issue i am just not clearly imagine is the influence of different
netmasks of the interfaces (or not answering icmp netmask-requests at
all) to the discovering process (ip_discover.tcl is quite complex...).
An interesting issue currently not included in the discoverer could be
a query about all interface-addresses via SNMP of the ipAddrTable.
This should also allow mapping of multiple interface addresses to one
node, supposing the box speaks snmp.
Erik
-- I can't remeber ever seen a )-: smiley before. But it really matches better than a (what the hell is this :-( smiley ;-)