CentOS - Aug 2007 - Forcing ifcfg-eth0 to use the same nic in multi nic machine

I am new to CentOS (coming from Fedora) and I really like it!

I am having difficulty getting one of my machines to boot and assign
the same designation of eth0 and eth1 to the same nics consistantly.

I have an MSI motherboard with 2 nics on the board.  Strangely enough
both nics report the same MAC address.  This is not an issue since I
use the computer to route between two different subnets so they don't
see each other on the network.

The problem comes when I reboot and they race against each other to
see which one will get to be eth0.  If they switch my routing dies and
I loose access with out rebooting and hoping they switch back or
switching the actual cables.

I have googled and searched my networking resources but all of the
fixes that I have found focus on using the MAC address to solve the
problem but in my case they are the same.

The only difference is that they use different drivers.  One is a
Realtek and other is a Marvell.

I have turned off Kudzu or they reconfigured every time a booted the
machine.  It only happens every once in a while but I need it to be
consistent.

Any suggestions or pushes in the right direction would be most appreciated.

Doug Coats

Doug Coats spake the following on 8/13/2007 11:38 AM:
> I am new to CentOS (coming from Fedora) and I really like it!
> 
> I am having difficulty getting one of my machines to boot and assign
> the same designation of eth0 and eth1 to the same nics consistantly.
> 
> I have an MSI motherboard with 2 nics on the board.  Strangely enough
> both nics report the same MAC address.  This is not an issue since I
> use the computer to route between two different subnets so they don't
> see each other on the network.
> 
> The problem comes when I reboot and they race against each other to
> see which one will get to be eth0.  If they switch my routing dies and
> I loose access with out rebooting and hoping they switch back or
> switching the actual cables.
> 
> I have googled and searched my networking resources but all of the
> fixes that I have found focus on using the MAC address to solve the
> problem but in my case they are the same.
> 
> The only difference is that they use different drivers.  One is a
> Realtek and other is a Marvell.
> 
> I have turned off Kudzu or they reconfigured every time a booted the
> machine.  It only happens every once in a while but I need it to be
> consistent.
> 
> Any suggestions or pushes in the right direction would be most appreciated.
> 
> Doug Coats
Is the boards bios up to date?
Do you have the proper alias entries in  /etc/modprobe.conf?


-- 

MailScanner is like deodorant...
You hope everybody uses it, and
you notice quickly if they don't!!!!

On 8/13/07, Doug Coats <dcoatshca at gmail.com>
wrote:
> I am new to CentOS (coming from Fedora) and I really like it!
>
> I am having difficulty getting one of my machines to boot and assign
> the same designation of eth0 and eth1 to the same nics consistantly.
>
> I have an MSI motherboard with 2 nics on the board.  Strangely enough
> both nics report the same MAC address.  This is not an issue since I
> use the computer to route between two different subnets so they don't
> see each other on the network.
>
> The problem comes when I reboot and they race against each other to
> see which one will get to be eth0.  If they switch my routing dies and
> I loose access with out rebooting and hoping they switch back or
> switching the actual cables.
>
> I have googled and searched my networking resources but all of the
> fixes that I have found focus on using the MAC address to solve the
> problem but in my case they are the same.
>
> The only difference is that they use different drivers.  One is a
> Realtek and other is a Marvell.
>
> I have turned off Kudzu or they reconfigured every time a booted the
> machine.  It only happens every once in a while but I need it to be
> consistent.
>
> Any suggestions or pushes in the right direction would be most appreciated.
>
> Doug Coats

As far as I understand networking, if you only have 1 MAC, you only
have 1 NIC.  You might have 2 connectors, but that seems really
strange.  It seems like this is some sort of undefined behavior.  I
think that having only 1 MAC really *IS* an issue.

Is it possible they are set up to do of link teaming or something like that?

My suggestion is to get another network card.

On Monday 13 August 2007 21:38, Doug Coats wrote:
<cut>
>
> Any suggestions or pushes in the right direction would be most appreciated.
Could you show us the the contents 
of /etc/sysconfig/network-scripts/ifcfg-eth0 
and /etc/sysconfig/network-scripts/ifcfg-eth1?


-- 
Todor Petkov
Bulgaria Online
tel. (+359 2) 91972
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<http://lists.centos.org/pipermail/centos/attachments/20070813/244b05ae/attachment-0001.sig>

> Could you show us the the contents
> of /etc/sysconfig/network-scripts/ifcfg-eth0
> and /etc/sysconfig/network-scripts/ifcfg-eth1?

ifcfg-eth0

DEVICE=eth0
BOOTPROTO=none
BROADCAST=xx.xx.xx.xx
IPADDR=xx.xx.xx.xx
NETMASK=xx.xx.xx.xx
NETWORK=xx.xx.xx.xx
GATEWAY=xx.xx.xx.xx
ONBOOT=yes
TYPE=Ethernet
USERCTL=no
PEERDNS=no

ifcfg-eth1

DEVICE=eth1
BROADCAST=xx.xx.xx.xx
IPADDR=xx.xx.xx.xx
NETMASK=xx.xx.xx.xx
NETWORK=xx.xx.xx.xx
BOOTPROTO=none
ONBOOT=yes
TYPE=Ethernet
USRCTL=no
PEERDNS=no

Both have static IP addresses.  One is external to a DSL connection
and the other Internal to a DMZ.

On 8/13/07, Doug Coats <dcoatshca at gmail.com>
wrote:
>
> I am having difficulty getting one of my machines to boot and assign
> the same designation of eth0 and eth1 to the same nics consistantly.
I posted something about this back on April 25,
Message-ID: <6bb609560704250801y6efe4ec1gbc513ea4f34d5721 at
mail.gmail.com>

What I said then was:

--------
The problem (as I recall) is that on each reboot the onboard NICs are
being discovered in a different order, so the MAC addresses in
/etc/sysconfig/hwconf don't match what is recorded in
/etc/sysconfig/network-scripts/ifcfg-eth*.

I believe what I had to do (I should have written it down, damn it)
was hand-edit /etc/sysconfig/hwconf to completely remove all the
references to the NICs (there may be more than one entry for each NIC
because of the flip-flopping), edit ifcfg-eth* to remove all
references to HWADDR, reboot again to let the cards be rediscovered,
and then again hand-edit ifcfg-eth* to insert HWADDR lines that match
the device assignments in the regenerated /etc/sysconfig/hwconf.

However, I may have at least the last step of that wrong.
--------

No one ever responded as to whether that solution worked for them.

On 13 August 2007, Johnny Hughes <johnny at centos.org> wrote:
<snip>
> How can 2 nics from different companies have the same mac address ....
> I don't think that is possible.
> each card needs a separate MAC address, at least the way I understand
> networking.
I agree with Doug and Johnny. My belief is that every NIC in the world
is supposed to have a unique MAC address. Although the OP has been using
these boards without problems, I suspect there is a defect on the
boards.

Sorry - I missed the original post - but I can think of one scenario where
two NICs have the same MAC address. However, it's extremely unusua.l.

The IEEE has managed the allocation of addresses for Ethernet, but IBM has
done the same for Token Ring. The two didn't talk, and so their allocated
address ranges overlap. Consequently there are a number of pairs of Token
Ring and Ethernet cards out there that have the same MAC address. Token
Ring/Ethernet bridges have to do some bit-fiddling to map TR MAC's to
Ethernet because of it.

That's probably not what the OP was enquiring about - unless he's
extremely
unfortunate and still has to deal with TR.

Best,

--- Les Bell, RHCE, CISSP
[http://www.lesbell.com.au]
Tel: +61 2 9451 1144
FreeWorldDialup: 800909