DHCPD : failover behavior problems.

Wed Jun 18 12:44:48 UTC 2008

Hi,

I'm writing here because i've got some trouble using the failover fonction in dhcpd.

Here is my configuration and context:

I have got 2 servers on the same LAN running the version 3.0.4 of dhcp server.

Here are the 2 configuration files:

            ==> The primary:

####################################

## configuration DHCPD de primary ##

####################################

ddns-update-style ad-hoc;

one-lease-per-client true;

authoritative;

failover peer "dhcp" {

         primary;

         address 172.16.21.34;

         port 519;

         peer address 172.16.21.37;

         peer port 520;

         mclt 120;

         split 128;

         max-response-delay 30;

         max-unacked-updates 10;

         load balance max seconds 3;

         }

log-facility local7;

local-address 172.16.21.34;

# include du fichier contenant les scopes. #

include "/etc/dhcp3/dhcpd.master";

include "/etc/dhcp3/dhcpd1.conf";

include "/etc/dhcp3/dhcpd2.conf";

            ==> The secondary :

            root at secondary:~# more /etc/dhcp3/dhcpd.conf

####################################

## configuration DHCPD de primary ##

####################################

ddns-update-style ad-hoc;

one-lease-per-client true;

authoritative;

default-lease-time 180;

max-lease-time 180;

failover peer "dhcp" {

         secondary;

         address 172.16.21.37;

         port 520;

         peer address 172.16.21.34;

         peer port 519;

         max-response-delay 30;

         max-unacked-updates 10;

         load balance max seconds 3;

         }

log-facility local7;

# include du fichier contenant les scopes. #

include "/etc/dhcp3/dhcpd.master";

include "/etc/dhcp3/dhcpd1.conf";

include "/etc/dhcp3/dhcpd2.conf";

            ==> And the common scope définition in the "/etc/dhcp3/dhcpd.master" file (for information, the "/etc/dhcp3/dhcpd1.conf" and "/etc/dhcp3/dhcpd2.conf" files are blank files).

default-lease-time 60;

max-lease-time 100;

subnet 172.16.21.32 netmask 255.255.255.224

            {

            -- not important subnet --

}

subnet 37.0.76.0 netmask 255.255.255.0 {

        option broadcast-address 37.0.76.255;

        option routers 37.0.76.254;

        pool {

                failover peer "dhcp";

                range 37.0.76.1 37.0.76.4;

                deny dynamic bootp clients;

                }

        group ADSL-Flore {

                host Fraise { hardware ethernet 00:14:7F:00:5D:A8;}

                host Agrumes { hardware ethernet 00:14:7F:00:5F:EB;}

                host Fleurs { hardware ethernet 00:14:7F:00:5D:49;}

                host Arbres { hardware ethernet 00:14:7F:00:5F:6E;}

                host Legumes { hardware ethernet 00:14:7F:00:5D:67;}

                host Poisson { hardware ethernet 00:14:7F:00:5F:FC;}

                host Mammifere { hardware ethernet 00:14:7F:00:5D:8B;}

                host laminestet { hardware ethernet 00:15:C5:C7:5C:B8;}

                filename "swupgrade_st2030m_ADSL-FLORE.inf";

                next-server 172.16.21.59;

                }

}

            ==> the dhcp clients reach those servers by 2 dhcp relay (one for each servers). 

When the 2 servers are up and the dhcp demon running, everything works fine. Problems occur when one server is down.

In fact I've got the 2 following behaviour:

==> If the server "secondary" fails, the server who stays up can assign the half of the scope to new clients and can renew all the already attributed addresses by server primary and secondary. It seams ok for me but after the mclt is expired (120 sec) the situation never changes, even after a night the primary server never gives an adresse of the other parts of the scope. When I read documentations and the man page of dhcp server I though that after this time the standing server should give all the adresse (the 2 half part of the subnet). I tried to put the server standing in the "partner-down" state by "omshell" and by putting the appropriate lines in the dhcpd.conf, but nothing changes.

What is more strange is that if one address have already been assign by the secondary server sometimes it worked (to deserve the whole scope) and other times not.

==> The second problem I encounter is the fact that when only one server is standing the state of the previously affected lease never comes back to "free". For instance, when there is one server left, and all addresses in the scope are assigned, if I disconnect one of the dhcp clients and connect another one, the adresse of the client I disconnected never became free and the new client can't get an IP.

I've tried this configuration with the version 3.0.4 and 4.1 of the dhcp server, my linux distribution is a debian with kernel "Linux primary 2.6.25-1-486"

Maybe I've miss something in the configuration or there some restrictions I didn't understand. 

So if someone can help me to resolve those problems, I would appreciate a lot.

Ty in advance.

Fabien

" Ce message et toutes les pieces jointes sont etablis a l'attention 
exclusive de ses destinataires et sont confidentiels.  L'internet ne 
permettant pas d'assurer l'integrite de ce message, le contenu de ce
message ne represente en aucun cas un engagement de la part de notre 
societe. Si vous recevez ce message par erreur, merci de le detruire 
et d'en avertir immediatement l'expediteur.

  This message contains privileged and confidential information.  Given 
that the internet does not allow us to make sure of the communication's
integrity,  the content of this message does not  represent in any case
a commitment from our compagny.  If you are not the intended recipient,  
please notify  us  immediately and delete  this  e-mail  (including any 
attachments) from your system. "
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/dhcp-users/attachments/20080618/2b6b8202/attachment.html>