confusion of installation process

[Michael DeHaan]

Klaus Steden wrote:
> I could have sworn Ray Van Dolson said this Thu, Jan 11, 2007 at 12:59:13PM -0800 ...
>   
>> I assume you are referring to nicdelay and linksleep?  Are there any others?
>> I've been trying to get this to work on a RHEL4 U4 system on a switch with
>> Spanning Tree enabled.  Even with long delays I am unable to get pump to get
>> an IP address.  I don't understand how, when the system is installed, ISC's
>> dhcp client is able to get an IP no problem (after a delay of course).
>>
>>     
> I've come to the conclusion that 'pump' sucks, and should be replaced with the
> ISC's dhcp client, if possible ... or made more robust.
>
> Klaus
>
> _______________________________________________
> Kickstart-list mailing list
> Kickstart-list at redhat.com
> https://www.redhat.com/mailman/listinfo/kickstart-list
>   

This is why "koan" (see http://cobbler.et.redhat.com), when preparing to 
re-provision a system, cracks open the initrd and embeds the kickstart 
file inside of it rather than hoping it can be successfully fetched. 
That may prove a useful trick and you are welcome to look at the source.

Unfortunately doing that for PXE boot scenarios (where you have to do 
that once per system if you are using a lot of kickstart templating) 
really doesn't scale at all, as it means keeping a lot of initrd's 
around and they are a bit slow to generate. It is very reasonable to do 
client-side though. So yes, I'm in the Anaconda could use better dhcp 
client support camp also.

--Michael