SUMMARY Erroneous duplicate IP addresses

From: Tessa Davis (
Date: Sat Mar 04 1995 - 06:26:10 CST

I posted the following request

> Advice needed, please
> Environment - Sparc 10 with two Ethernet cards running Solaris 2.3
> Problem - Keeps reporting its own second card as being a different
> machine with duplicate address. Fills /var/adm/messages with
> hostname unix: IP: Hardware address 'xx:xx:xx:xx:xx:xx' trying to be
> our address !
> (Real names changed for security reasons)
> I have tried specifically changing hardware address of second card
> but has no effect.

With hindsight, I should have further explained that one ethernet card is
attached to a big LAN and uses a default-style IP address. The other card
is attached to a WAN via ISDN bridges and uses a "real" IP address. I.E.
the two addresses are also in different classes.

Several respondants wrongly assumed that I had both cards attached to the
same physical LAN. This would indeed produce a similar effect, because
the machine may see an echo of itself.

One respondant had a similar problem in an ISDN environment
From: heldesc!colinm (Colin Maytum) colinm@heldesc.uucp
.... only happens on Solaris 2.3. We had 4 sparc 10's running
   2.3 with only one ethernet card doing this. I suspected the problem was due
   to the ISDN line. When it dialed up we got these errors.....
   .... We rebridged the network and they went away.
   it may not be the 2 cards causing the problem, but a bug in the O.S.
This is similar to our environment. The problem does not occur on a sparc 10
running 4.1.3, even though it is physically configured identically and linked
to another WAN via the same type of ISDN bridge. The message does indeed
always coincide with a dial-up.

A disturbing but plausible explanation was received from
as follows:

You are using malfunctioning bridges in your network.

I strongly advise against the use of bridges in an IP network.
IP networks should be subnetted if they grow to large.
The Solaris Sysadmin answerbook gives adequate explanation on how to go about

The second cause is the most probable since a lot of so called
network consultants advertise the use of bridges. Believe me : they are wrong !
Bridges only have to be used with inferior networking protocols that do not
allow routing : Netbeui, ...
Get rid of those protocols, ban them of your network, It will help you sleep
better at night.

Koen Peeters

Other suggestions :

There is some sort of loop in your network that
is causing this machine to see its own packets echoed back.

Someone had put the wrong netmasks on a gateway machine and it got very
   confused (masking the same address for class B & C)

I've seen that in a switched environment, specifically a SparcCenter 2000
(Solaris 2.3) on a Grand Junction FastSwitch 10/100 switched hub. It happened
when we enabled monitoring. Went away as soon as we disabled monitoring.
I guess all its own packets were being echoed back to
itself and it got confused.

as far as I know all solaris 2.3 installations do this even
with just one ethernet

you have a problem with something responding to
all arp requests -- when an interface comes up, it arps itself
to look for a duplicate IP. watch the exchange using snoop
(do this on another machine) and see who answers the ARP request
from the machine in question.

SUMMARY As I said before, the problem does not occur with a 4.1.3 system
using identical bridges. It seems to be a interaction between Solaris 2.3
and the bridges. Still not sure if one or both has a bug.
Unfortunately I have 100 of these ISDN bridges in my WAN, and can't afford
to replace them with routers at the moment.
I will probably try to put a router between the WAN and the LAN.

Many thanks for all replies

Tessa Davis

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:17 CDT