fixed-address reservations not being honored

Simon Hobson dhcp1 at thehobsons.co.uk
Thu Sep 28 18:44:40 UTC 2006


Aaron Bennett wrote:

>I've got a wierd problem.

Not really, if you understand both ends of the system.

>I have a Windows XP host which received a
>lease for 10.6.65.110; later I decided I'd like to move it to a reserved
>address, 10.6.65.125.  I've added the reservation to the dhcp.conf file
>on both of the failover nodes, but it ipconfig /renew kept getting
>10.6.65.110.  So, taking an extra step, I did ipconfig /release, deleted
>the lease from both nodes leases file and did an ipconfig /renew --
>still got 10.6.65.110.

Windows seems particularly adept at keeping addresses ! Even though 
you've done a release, I believe it still requests the address it 
last had ...

>Taking it even further, I excluded 10.6.65.110
>from the range and STILL, I keep getting 10.6.65.110!
>
>Any thoughts?

See below ... BTW, you did restart the server didn't you ? And shut 
down both servers before editing the leases file ? And removed ALL 
versions of the lease from both files ?

>   Here's the relevant snips from the config files....
>
>subnet 10.6.65.0 netmask 255.255.255.0 {
>                 authoritative;
>                 pool {
>                         failover peer "failover" ;
>                         range 10.6.65.51 10.6.65.109;
>                         range 10.6.65.111 10.6.65.254;
>                         option routers  10.6.65.1;
>                         option domain-name-servers      140.232.1.32,
>140.232.1.33;
>                         option domain-name      clarku.edu;
>                         option netbios-name-servers     10.1.1.10,
>140.232.1.13;
>                         option netbios-node-type        8;
>                         option default-lease-time       691200;
>                         deny dynamic bootp clients;
>                  }
>          }
>
>host ABENNETTD600 {
>         hardware ethernet 00:0f:1f:c8:91:73;
>         fixed-address 10.6.65.125;
>};
>
>and /var/log/messages shows:
>
>Sep 28 13:42:47 dhcptest1 dhcpd: DHCPREQUEST for 10.6.65.110 from
>00:0f:1f:c8:91:73 (ABENNETTD600) via eth0
>Sep 28 13:42:47 dhcptest1 dhcpd: DHCPACK on 10.6.65.110 to
>00:0f:1f:c8:91:73 (ABENNETTD600) via eth0
>       
>how could this be?  That pool's range doesn't include the address it's
>getting, I'm confused (of course).

The client requests a particular address, and although it is not in 
any pool, it is still valid for the subnet.

It's come up before when changing pools around, you may have to 
explicitly deny booting from a range to force clients to be NACKed. 
Not sure why it ACKed it though, my (dodgy) memory tells me that it 
simply ignored requests like this (valid for subnet but not in a 
pool).

   pool {
     range 10.6.65.110 ;
     deny booting ;
   }




More information about the dhcp-users mailing list