Rép. : auto partner-down

[Pablo Gaston Milano]

Yes, but what that statement works in a different way:
If the "split" hash is being used, each client request is hashed, and only 
one server sends a response, based on the hash value. If, after "load 
balance max seconds", the server which should respond did not do it, the 
other does it. This feature was thought to address the case in which a 
server is partially down, so it can communicate with the peer, but is not 
assigning leases.
Unfortunately, this statement does not help in the scenario I presented.





"Serveur-Faucon Surveillance" <SrvFaucon at cslaval.qc.ca> 
Enviado por: dhcp-users-bounce at isc.org
22/11/2006 11:53 a.m.
Por favor, responda a
dhcp-users at isc.org


Para
<dhcp-users at isc.org>
cc

Asunto
Rép. : auto partner-down






Do you have the "load balance max seconds" statement in your dhcpd.master 
file?



---------------------------------------------------
Alexandre Racine - Gardien Virtuel - Sécurité Informatique 
www.gardienvirtuel.com
Montréal, Québec, Canada

>>> pgmilano at quilmes.com.ar 2006-11-21 15:51:52 >>>
I am using DHCP with Failover. In my case, the lease time is set to twelve 

hours. Unfortunately, the IP ranges are used in more than half the amount 
of IP addresses and it is really hard to change that.
Some times happens that, during the weekend, one of the servers goes down, 

making the survival server to enter the "communications-interrupted" 
state. On monday, when users start to power on their workstations, the 
survival server quickly reachs its 50% of the total pool size, rejecting 
requests of all other users. The solution here, is to go quickly and put 
the survival server into "partner-down" state. But unfortunately, users 
ofthen realize of the problem before administrators do.
In order to avoid that, I find really usefull to be able to chose a time 
period, after which a server in "communications-interupted" state change 
into "partner-down" state. This period is specified in the ietf drat as 
"safe period".
In some other previous thread, some users proposed a new feature called 
"auto-partner-down-time". I totally agree with that. It would be good to 
the administrator to have the choice to configure this.