ethernet bonding question

[Eric Sisler]

On Wed, 2004-08-11 at 17:36, Michael Rubin wrote:

> I just set up a round-robin ethernet bond with 2 nics, it seems to be
> working (both cards sending and receiving).  /proc/net/bonding/bond0 gives
> the proper hardware addresses for eth0 and eth1 (from
> /etc/sysconfig/network-scripts/ifcfg-ethX I assume, where the hardware
> addresses are set correctly), however, /sbin/ifconfig displays the same
> hardware address for bond0, eth0 and eth1 (eth0's).  Should I be worried
> that the hardware address for eth1 is displayed incorrectly, or is that just
> a side-effect of the bond?

It's a side-effect of the bond, no need to worry.

> On a side note, the RX & TX numbers (from /sbin/ifconfig) for eth0 and eth1
> seem to be somewhat unbalanced (eth1 receiving 5 times as much as eth0, TX
> is exactly the same for both).  Is that a switch configuration issue or a
> server configuration issue?

I recently setup bonding as well, so I'm not sure.  Incoming packets
seem to favor one interface here as well, outgoing are nearly the same.
My theory at this point is that incoming traffic is only using one
interface (via round-robin) while outgoing traffic uses both.  I need to
re-read the bonding documentation again to see if it provides any more
details.  What kind of switches do you have and how are they
configured?  I have some older Cisco 3500-XL's here, configured for port
grouping.
 
What's mildly entertaining is pulling the plug on one of the cables
without interrupting the data flow.  ;-)  You'll see messages to that
effect in /var/log/messages and /proc/net/bonding/bond0 will increment
the failure count.

-Eric

-- 

Eric Sisler <esisler at westminster.lib.co.us>
Library Applications Specialist
Westminster Public Library
Westminster, CO USA

Linux - Don't fear the Penguin.
Want to know what we use Linux for?
Visit http://gromit.westminster.lib.co.us/linux