Comments on Booting On PXE And On A Customized Debian System

That sounds wrong. I get the same error message on a very dumb switch. I can mount the NFS share on other machines, just not as nfsroot when PXE booting.

hi this how-to is very good and i learn how to use PXE whit this.

but it has some typos and some errors.

in chapter 3 in the lan configuration, if you use dhcp there the boot crash...

 and in chappter 6 where say

cp /pxeroot/initrc.img ./

should say cp /pxeroot/initrd.img ./

( cp /pxeroot/initrD.img ./ )

 

thanks for this how to 

alternative good approach here...

http://www.vercot.com/~serva/DebianPXE1.html

actually the link is
http://www.vercot.com/~serva/howto/DebianPXE1.html

Yes, it looks good, but ... it never worked for me!

I installed it several times (debian etch 4.0 r3), I've been searching a damn clue everywhere, BUT :

nfs: server not responding, still trying

until the end of eons... And no answer.

Lots of people suffered with this, just check Google, but I found no answer for my particular case :

- network cable OK;

- switch changed 5 times, with different HW;

- nfs is working (??? whatever, I can mount it by hand);

- DHCP OK (client gets valid address );

- TFTP looks good too;

No idea at all of what could trick.

Cheers

This sounds similar to two problems I've had before.

The first goes something like this:

• You PXE boot and DHCP works
• You start the installer
• It times out waiting for a network service (DHCP, NFS, etc)

The problem is Spanning Tree or other network negotiations. When the second attempt to connect to the network happens the switch port has to do it's Spanning Tree negotiation. This can take 30-60 seconds and the service waiting to get a reply from the network times out. The fix was to enable "port fast" on the switch port that the server was connected to; this skips the negotiations that happen. Cisco calls the option "port fast" though it may be different for other vendors.

The second problem I had was a box that had two NICs. PXE boot detected one NIC as eth0 but the kernel detected the other as eth0. The problem has to do with how the BIOS presents the NICs to the kernel and how the kernel determines which NIC is first. The solution was to physically switch ports between PXE boot and the installer doing DHCP. Another option is to disable one NIC in the BIOS. This will have consequences if you re-enable the NIC after you install since the kernel will still detect the other NIC as eth0.

Solution to "nfs: server not responding, still trying" for me was disabling network configuration in /pxeroot/etc/network/interfaces (network is allready configured at that time).

Also if you build custom kernel, don't forget to change "BOOT=nfs" in initramfs.conf, before you create initrd image (e.g. update-initramfs -c -k vmlinuz-2.6.25.4-custom ).

Great tutorial by the way.