What is the proper method to restart two node failover-dhcp services?

Glenn Satchell glenn.satchell at uniq.com.au
Fri Dec 2 01:40:14 UTC 2011

Unless you know there is serious corruption, removing the leases file is
almost never required, and is actually a bad thing to do because now yor
dhcp serve is out of sync with the clients who think they have various
leases. the ping-first type checks are not all that useful in these days
of personal firewalls on clients. when the client restarts, the first
thing it does is request a lease synchronisation from the other server.

The logic behind step b) is that if you define new subnets, then restart
the primary it will send all the new subnet info to the secondary which
won't know aout those subnets.

If you have a large lease file, then you may need to allow enough time
between restarting one server and the other for the lease synchronisation
step to complete. You can follow this in the log files to see how long it
takes. Usually it is quick, but a 1-2 minute pause may help here if you
have lots of leases.

5 minute leases are very short, and it could be that this close enough to
the timing period between the restarts. Are you able to make them longer
(even 10 minutes might make a difference)?


> This is what I remember reading on this list:
> a) Bringing both servers back into sync: Move or rename the leases
> file on the secondary. Restart dhcpd on the secondary.
> b) Restarting after making config changes: Restart the secondary
> first. When that is done, restart the primary.
> Hope this helps.
> cv
> On Thu, Dec 1, 2011 at 11:52 AM, Marion Bogdanov
> <marion.bogdanov at gmail.com> wrote:
>> Kind folks,
>> We are experiencing some issues with our two failover configured DHCP
>> servers.
>> In particular, for short leases (300 seconds), which is predominantly
>> wireless users, the users are experiencing problems that let us to
>> discovering that the dhcp server is issuing duplicate IPs.
>> We are lead to believe that this may be caused by our method used to
>> restart
>> the services, after we have made changes to the scopes files, which is:
>> root at dhcp1#   /usr/sbin/dhcpdctl stop; /usr/sbin/dhcpdctl start;
>> followed by:
>> root at dhcp2#   /usr/sbin/dhcpdctl stop; /usr/sbin/dhcpdctl start;
>> Thanks in advance for your much appreciated input.
>> -Marion

More information about the dhcp-users mailing list