Hi. I have been running a boot server for many years now, and I have never had a problem like this before. My boot server has 2 NICs: one on our network and one on a private network (switch). When I want to install the OS onto a new client, I connect the client to the private switch, register its MAC address with my boot server, and do a PXE boot from the client. The client talks to the server and accesses a "default" file located in the /tftpboot area of the server. This file, called "default" actually brings up a menu and allows the user to choose a distro, and times out after 10 seconds to the first choice. Here's a snippet:

<code>
prompt 1
default 1
timeout 100 # 10 seconds
display msgs/boot.msg
.
.
.

# 1st option (user presses 1)
label 1
kernel /images/pxeboot/fc8-i386/vmlinuz
ipappend 2
append ks=nfs:192.168.0.1:/repo/fc8-i386/ks/ks8.cfg ksdevice=link initrd=/images/pxeboot/fc8-i386/initrd.img ramdisk_size=10000
.
.
.
.
# 8th option (user presses 8)
label 8
kernel /images/pxeboot/fc10-i386/vmlinuz
ipappend 2
append ks=nfs:192.168.0.1:/repo/fc10-i386/ks/ks10.cfg ksdevice=link initrd=/images/pxeboot/fc10-i386/initrd.img ramdisk_size=10000
</code>

To ensure that /repo is readable by NFS, I added this line in /etc/exports:

<code>
/repo 192.168.0.0/255.0.0.0(rw,no_root_squash,sync)
</code>


It gets to this screen, fine, and if I choose any other install (Fc3-Fc8), it works fine. I haven't changed the way I do things for the Fc9 or Fc10 installs. However, I have recently tried adding both Fc9 and Fc10 (downloaded the rpms into a local YUM repo, downloaded the iso file and mounted it locally, created a kickstart file for each in /repo/fcX-i386/ks/ksX.cfg, etc.) For some reason, when I try a Fc9 or Fc10 install, it runs anaconda, but says:

<text>
Unable to Download the Kickstart file, Please modify the Kickstart paramater below or press cancel to proceed as an interactive installation.
</text>


So I know it's not even getting to the kickstart file itself. I've seen some similar posts before: http://www.linuxquestions.org/questi...roblem-644105/
and I've looked through the release notes for changes in Fedora 9 and 10, but can't seem to find any explanation for why it can't use nfs to properly get the kickstart file, since it works for all the previous installations. At first, I thought that perhaps it was a kernel bug, and the kernel that I was using for Fc9 needed to be fixed. However, after getting the same results for Fc10, I think it's too coincidental and must be a change in how anaconda uses nfs or something. Except I can't find any documentation to support this notion anywhere.

If anyone knows what "changed" between Fedora 8 and Fedora 9/10 in this regard, please let me know so I can try to alter my approach, and see if it works.

Thank you in advance!
-Brendon

Between 8 and 9 nothing as far as i know .but 9 and 10 there is a change in how network-manager see multiple networks during the install . I was going to suggest reading the 10 release notes , but you have already .
But seeing that it is a problem with 9 and 10 , you got me .

Yes, I saw that, however, I don't think that network manager is having trouble identifying a network... The page says:

"Anaconda is now using NetworkManager for configuration of network interfaces during installation. The main network interface configuration screen in Anaconda has been removed. Users are only prompted for network configuration details if they are necessary during installation. The settings used during installation are then written to the system."

I don't see how that would prevent a kickstart file from being downloaded via NFS, would you?

Has anyone had any experience with kickstart over a private network for any distros of Fedora?