DHCP failover issue

[Jesper Nørgaard]

Hi,

I have just taken over a two node DHCP setup, running in failover configuration.
We have just recently discovered issues with some clients not getting an IP address.
What I see in the logs:
Server 1 logs: load balance to peer
Server 2 logs: peer holds all free leases

For the pool in question, server 2 shows this:
sc-dhcp-02 dhcpd: balancing pool b23e820 x.x.x.x/21  total 2000  free 2000  backup 0  lts -1000  max-own (+/-)200

Server 1 shows nothing for this pool.

Much to my surprise I discovered that the dhcpd.master configuration files are not at all identical on the two servers.
For the address pool in the example:
Server 1:

subnet x.x.x.x netmask 255.255.240.0 {
                             option dhcp-parameter-request-list 1,3,6,15,42,43,54,128,201,230,232;
#        default-lease-time 900;
#        max-lease-time 1000;
        option routers                  x.x.x.x;
        option subnet-mask              255.255.240.0;
        option domain-name-servers      x.x.x.x;
        option AgentID 10.23.16.1; # Option 54
        option NTP-server 88.83.68.3; # option 42
#        pool {
#                failover peer "dhcp-failover";
                range 10.23.16.16 10.23.31.254;
#                deny dynamic bootp clients;
#        }
}

Server 2:

subnet x.x.x.x netmask 255.255.240.0 {
option dhcp-parameter-request-list 1,3,6,15,42,43,54,128,201,230,232;
#        default-lease-time 900;
#        max-lease-time 1000;
        option routers                  x.x.x.x;
        option subnet-mask              255.255.240.0;
        option domain-name-servers      x.x.x.x;
        option AgentID x.x.x.x; # Option 54
        option NTP-server x.x.x.x; # option 42
        pool {
                failover peer "dhcp-failover";
                range 10.23.16.16 10.23.31.254;
                deny dynamic bootp clients;
        }
}

As you can see, server 2 does have a "failover peer" statement, server 1 does not.
The configuration above is similar to a lot of other pools in the master config.
I have no idea why anyone would configure the servers like this, but it's obviously not working.

Any idea on how to get addresses to the clients?

Thanks!


Med venlig hilsen/ Best regards

Jesper Nørgaard
IT System Administrator

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/dhcp-hackers/attachments/20120712/c82316c0/attachment.html>