Syslinux - Sep 2006 - PXELINUX booting delays with dual Ethernet ports

We're using PXELINUX for bootstrapping the network install our Linux cluster
nodes.  We just got some new IBM 326m dual-Opteron servers with Broadcom
NetXtreme BCM5780 dual-Gigabit Ethernet ports.

These dual-Ethernet servers work in an annoying way with PXELINUX when we
configure for booting from the local hard disk as the PXE-client downloads
this file from the DHCP/TFTP server:

label harddisk
   localboot 0

The first Broadcom Ethernet port correctly does DHCP and TFTP and picks up
that it must boot from the local hard disk (as printed on the screen).

Problem: In stead of actually booting from hard disk, the Broadcom
hardware then goes on to PXE-boot the second Ethernet port (which is also
connected to the Ethernet switch).  After a minute or two (when the second
port receives no response from the DHCP server), finally the desired
hard disk boot will proceed as expected.  We would like to get rid of
this superfluous booting delay.

Question:  Is there any way to make PXELINUX bypass the PXE booting of
the second Ethernet port and proceed immediately to booting from local
hard disk ?

Additional details:
Our DHCP/TFTP server runs RedHat Linux RHEL4 Update 4 with syslinux-2.11-1.
Our PXE-booting Linux cluster nodes have both Ethernet ports connected
to the same switched network, since we at a later stage configure Ethernet
bonding of the two ports eth0 and eth1 in order to achieve load balancing
and increased bandwidth.

Thanks a lot,
Ole

-- 
Ole Holm Nielsen
Department of Physics, Technical University of Denmark

On Fri, Sep 01, 2006 at 02:46:02PM +0200, Ole Holm Nielsen
wrote:
> We're using PXELINUX for bootstrapping the network install our Linux
cluster
> nodes.  We just got some new IBM 326m dual-Opteron servers with Broadcom
> NetXtreme BCM5780 dual-Gigabit Ethernet ports.
> 
> These dual-Ethernet servers work in an annoying way with PXELINUX when we
> configure for booting from the local hard disk as the PXE-client downloads
> this file from the DHCP/TFTP server:
> 
> label harddisk
>    localboot 0
That tells the bios to continue to the next device. If you want to force
harddisk boot, use chain.c32 instead. (chain hd0, i believe.)

-- 
lfr
0/0
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<http://www.zytor.com/pipermail/syslinux/attachments/20060901/308d53f2/attachment.sig>

Luciano Miguel Ferreira Rocha wrote:
>> We're using PXELINUX for bootstrapping the network install our
Linux cluster
>> nodes.  We just got some new IBM 326m dual-Opteron servers with
Broadcom
>> NetXtreme BCM5780 dual-Gigabit Ethernet ports.
>> 
>> These dual-Ethernet servers work in an annoying way with PXELINUX when
we
>> configure for booting from the local hard disk as the PXE-client
downloads
>> this file from the DHCP/TFTP server:
>> 
>> label harddisk
>>    localboot 0
> 
> That tells the bios to continue to the next device. If you want to force
> harddisk boot, use chain.c32 instead. (chain hd0, i believe.)
Interesting observation !  This, however, contradicts the FAQ
http://syslinux.zytor.com/faq.php which says:
   On PXELINUX, specifying "LOCALBOOT 0" instead of a
"KERNEL" option means
   invoking this particular label will cause a local disk boot instead of
   booting a kernel.

Could you kindly specify in exact detail the SYSLINUX configuration
file for forcing an immediate boot from hard disk (say, Linux /dev/sda)
in stead of the above "LOCALBOOT 0" ?  I find the FAQ too dense for my
(lack of) understanding.

Thanks a lot,
Ole