[K12OSN] Tossing around an idea...need some input

[John Baillie]

David Trask wrote:

>"Support list for opensource software in schools." <k12osn at redhat.com> on
>Saturday, August 13, 2005 at 9:36 PM +0000 wrote:
>  
>
>>Yes:
>>
>>include "filename";
>>will do what you expect.
>>More than you want to know is at: http://www.bind9.net/manuals-dhcp
>>Note that the involved process for dhcp failover is only for the
>>case where you need overlapping dynamic ranges - and it is partly
>>designed to keep giving the same IP address to the same requester
>>even if one of the dhcp servers goes down.  In the multi-ltsp server
>>case that really doesn't matter and you really don't want the rest
>>of the dhcp-provided info to be the same because you need to send
>>them to a new serv
>>    
>>
>
>this idea works except it's possible for one machine to be faster than the
>other, thus getting all the glory...isn't it?  Hence the load balance
>never happens when it comes to terminals.  OR  if you did a small range
>of, say, 30 addresses on one server.....it's entirely possible for
>standalone workstations (windows machines, ugh) to take up all 30 of those
>addresses....thus the terminal server never even gets used....right?
>
>David N. Trask
>Technology Teacher/Coordinator
>Vassalboro Community School
>dtrask at vcsvikings.org
>(207)923-3100
>
>_______________________________________________
>K12OSN mailing list
>K12OSN at redhat.com
>https://www.redhat.com/mailman/listinfo/k12osn
>For more info see <http://www.k12os.org>
>
>  
>
We experimented using three k12servers and one took precedence even when 
we put a heavy load on it. We also ended up with some gray screens with 
the X. One server won by a factor of about 3 to 1.  Also it seemed to 
take noticeably longer from power up to log in screen when we had 
multiple servers running.

It is quite possible that people could log on first thing in the morning 
and not run any apps and the majority of requests are met by one server. 
Then later when people start working the server will become bogged down.

Something else to consider is this, even with gig to the switch lets say 
that large file transfers are taking place on server1 at the same time a 
group of terminals are booted up, it is quite possible that server1 
would not hand out the os because it is slower to respond.

I'm not sure what the status is on setting up a chooser on CentOS but if 
it works we will probably go with that for the time being. Option two is 
hard coding mac addresses either via dhcp or IP tables.

John