do the funky eth...

[Nick Estes]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Try putting the iproute2 tools on there, they'll at least help you
troubleshoot a little more.  For example, when you take down an interface
with ifconfig, it still has the ip address in the kernel, with iproute2
you can see those and pull out the bad IP.  It also gives you better
control over the arp table too.

	--Nick

On Fri, 2 Mar 2001, der.hans wrote:

> moin, moin,
>
> running into a new eth prob where I'm consulting.
>
> Had a box the other day that had two ethernet cards, each with its own IP
> addy. The addies happened to be on the same subnet and be hooked in to the
> same switch. ifconfig and route were setup correctly. After shutting down
> the second interface and disconnecting the cable the box kept answering
> for it. First IP still worked too, so it wasn't just me getting things
> backwards :).
>
> Anyway, I would think that shouldn't work for a couple of reasons. First
> the IP stack shouldn't allow it. Second the switch shouldn't be sending
> packets for the second IP addy to the MAC addy of the primary card.
>
> Today, ran into another prob. Needed to move a service from one box to
> another. Had the second box ready. Instead of changing DNS and waiting for
> updates, which would've not worked well with that service, I just swapped
> IP addies. Took the old box off the air, changed the IP on the new
> box, then brought the old box up on the IP from the new box. That works
> fine, except they both continue to answer to the old IP addies, e.g. both
> boxen think they have both IP addies. Networking is setup correctly. The
> incorrect IP doesn't show up in ifconfig or route.
>
> On the old box the arp entry for the old IP is munged. arp -d won't kill
> the entry. Staticly setting the arp entry to the correct MAC addy doesn't
> fix the problem.
>
> On the new box the old IP isn't even in the arp table.
>
> Doing much of this from the consoles. Everything works fine from other
> boxen, e.g. service is only running on one box now and the correct box
> responds.
>
> Don't know if the following matters, but here it is anyway :). Both boxen
> are running RH 6.2, but have the errata installed. The old box is running
> 2.2.16 and the new box is running 2.2.17. The service is something
> develled in house, but it's just a tcp daemon and shouldn't affect the
> networking.
>
> Rebooting the old box didn't help.
>
> Anybody have any ideas on what's wrong and how to fix it?
>
> ciao,
>
> der.hans
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.4 (GNU/Linux)
Comment: mmm... gpg is yummy...

iD8DBQE6oBP0v+hjYTGg7s4RAmjbAJ48yX0OdS0F7K8e4FkOwzQnODWrpACfVyyw
detC9nEiUS3TBWuvGwGEDQM=
=YXRY
-----END PGP SIGNATURE-----