Hi,

[log in to unmask] wrote:

> 
>
> > As you see, we're accessing the kickstart files via NFS rather than 
> > via http:// - I'm not sure how important the difference might be.
>
>I found the difference is important.  I've ended up using the NFS
>/kickstart/a.b.c.d-kickstart method because of it's flexibility:
>I have a script that automates DHCP config and the creation of
>various "classes" (eg compute node, storage server, desktop)
>kickstart config files from a master/template kickstart config file.
>
>
>  
>
For our installations, we also have scripts that manage DHCP config and 
build kickstart
from templates for the machines.
But we swtich to "ks=file:/tmp/ks.cfg" as our script mounts the initrd 
(-o loop) and adds the
kickstart file inside it. Of course we have to maintain a initrd for 
each machine, but it prevent
another network access. In fact this method is used to have several 
scripts that are executed
at %pre and at %post (--nochroot) without additionnal network connection.

It is mainly usefull if you have other files than the kickstart, but in 
case of many machine
installations at the same time and if the kickstart is on the same 
machine than the RPMs
the NFS server might be very busy, so it improves speedup.

--
Yannick Perret
CC-IN2P3