Bonding with VLAN's, was Re: [Beowulf] Re: Beowulf Digest, Vol 18, Issue 9

Tim Mattox tmattox at gmail.com
Tue Aug 9 05:47:02 PDT 2005


Hello Eduardo,
I don't have any experience with using VLANs in combination with bonding,
however, I do recall various bugs & fixes related to VLAN support
being discussed
on the bonding development mailing list this year.  I don't recall the
final outcome.

I'd suggest you try to configure your network without the use of VLANs.
If you don't have a second standalone switch to use as the second channel,
you must configure the switch to use
"trunking/Etherchannel/whatever-it's-called-by-that-vendor".

The two NICs in each node will have identical MAC addresses.
Unless you use the physically disjoint switch per channel connection method,
you must configure the switch for that non-standard situation, or it
will get confused.
As I said, I don't have experience with VLANs, but I would not have
any confidence
that a switch split in half using VLAN technology would not still get
confused by
the bonding's duplicated MAC addresses.

As for why you see the HWaddr (MAC) being different between the two NICs on
another node, that should not be.  Something is misconfigured, or the network
cards in the other machine don't support changing their MAC address.
You should make sure you have installed a recent version of ifenslave.
The ifenslave.c source comes with the kernel source tree in
Documentation/networking/

Good luck.

On 8/8/05, Eduardo N. Asada <asada at dsee.fee.unicamp.br> wrote:
> Hi Tim,
> I have done exactly the same way it is described in the bonding documentation.
> The switch is a 3com Superstack II with Vlan and it is configured as a
> untagged vlan mode, so that it is splitted into two vlans. I have noticed
> that on one node the two network cards (eth0 and eth1) are bonded and the
> HWaddr are the same, however on the other node the HWaddr are different! When
> I add manually the second card, it becames very slow. These are situations
> that aren't covered in the manual...
> 
> --
> Eduardo N. Asada
> http://www.dsee.fee.unicamp.br/~asada
> email: asada at dsee.fee.unicamp.br
>          easada at ieee.org

-- 
Tim Mattox - tmattox at gmail.com
  http://homepage.mac.com/tmattox/
    I'm a bright... http://www.the-brights.net/




More information about the Beowulf mailing list