Bonded head nodes

[ivan at sixfold.com]

Hi Rob,

The Multicast extension to TFTP described in the PXE standard has no
mention of the concept of bonded ethernet channels. I am not familiar
enough if ethernet bonding to say with certainty, but my initial guess
would be that the ROM-driven stage of the PXE boot process would most
definitely not support this due to the fact that a single card on a
given client node is driven by the PXE code in the BIOS.

If, however, bonded ethernet were supported in the second stage boot,
which is driven by PXELinux or a similar bit of client code, there's
no reason a slight alteration to the server and client would not
easily support this type of transfer in the unicast case. Mixing the
concepts of multicast and ethernet channel binding seem odd and I'm
not entirely sure if it would result in any kind of performance
enhancement. If someone out there is working on improvements like
thisd, I would like to hear about it. We are very interested in
participating in the development of higher performance boot client
code for Linux clusters.

Thanks,
Ivan...




On Fri, 8 Nov 2002 robnash at rogers.com wrote:

> Dan, How does one bootup the nodes in bonded mode using the NIC's onboard Bootprom? 
> > 
> > From: Daniel Pfenniger <daniel.pfenniger at obs.unige.ch>
> > Date: 2002/11/06 Wed PM 01:19:49 EST
> > To: robnash at rogers.com
> > CC: beowulf at beowulf.org
> > Subject: Bonded head nodes
> > 
> > Hi Rob, 
> > 
> > It is possible, but the head nodes require 3 NICS, 
> > the third NIC is for the world. 
> > 
> > We have a Beo-cluster fully dual bonded, with two head nodes. 
> > Here are the relevant files (RH7.3):
> > 
> > /etc/modules.conf
> > --------------------------
> > alias eth0 tulip
> > alias eth1 tulip
> > alias eth2 tulip
> > alias bond0 bonding
> > ...
> > --------------------------
> > 
> > /etc/sysconfig/network-scripts/ifcfg-bond0
> > --------------------------
> > DEVICE=bond0
> > IPADDR=192.168.1.100
> > NETMASK=255.255.255.0
> > NETWORK=192.168.1.0
> > BROADCAST=192.168.1.255
> > ONBOOT=yes
> > USERCTL=no
> > BOOTPROTO=none
> > --------------------------
> > 
> > /etc/sysconfig/network-scripts/ifcfg-eth0
> > --------------------------
> > DEVICE=eth0
> > BOOTPROTO=static
> > BROADCAST=xxx.xxx.xxx.xxx
> > IPADDR=xxx.xxx.xxx.xxx
> > NETMASK=255.255.255.0
> > NETWORK=xxx.xxx.xxx.0
> > ONBOOT=yes
> > --------------------------
> > 
> > /etc/sysconfig/network-scripts/ifcfg-eth1
> > --------------------------
> > DEVICE=eth1
> > USERCTL=no
> > ONBOOT=yes
> > MASTER=bond0
> > SLAVE=yes
> > BOOTPROTO=none
> > --------------------------
> > 
> > /etc/sysconfig/network-scripts/ifcfg-eth2
> > --------------------------
> > DEVICE=eth2
> > USERCTL=no
> > ONBOOT=yes
> > MASTER=bond0
> > SLAVE=yes
> > BOOTPROTO=none
> > --------------------------
> > 
> > 	Dan
> > 
> > robnash at rogers.com wrote:
> > > 
> > > Hello Everyone,
> > > 
> > > Fist off I would like to thank everyone for the valuable input regarding the "Two heads are better that one" question.
> > > 
> > > I have another question. I have been doing quite a lot of reading on how to bond Ethernet interfaces. In all my reading I have yet to come across literature that describes booting nodes in bonded mode. Is this possible? I have been hacking around with PXE and it seems to me it's not doable due to the fact that the boot process happens in two phases? Does anyone out there know the steps involved in get a fully bonded cluster up and running using the NIC?s prom as the initiator?
> > > 
> > > I have the INTEL 10/100/1000MX copper cards.
> > > 
> > > Thank you in advance,
> > 
> 
> 1
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 


----------------------------------------------------------------------
Ivan Pulleyn            2000 West Carroll Ave Suite 303A
Sixfold Technology      Chicago, IL 60612-1600
ivan at sixfold.com        voice: (312) 421-0834    fax: (312) 421-0388
----------------------------------------------------------------------