DHCPFailoverTroubleShooting

1. dhcpd server processes must be able to communicate with each other. That means adding a rule on the master system that applies to both system allowing the dhcpd ports. I suggest creating a alias for this. In the future we need to add these rules and ports to the default dhcp server filter rules we already generate. e.g. create a self_ip alias which holds all the carp cluster local IP addresses including the carp addresses themself. Then create a allow rule on each interface that allow traffic from the self_ip to the self_ip on all protocols. That should cover it.

2. The system time on both carp systems need to be within 90 seconds of drift of each other otherwise the dhcpd processes will not communicate because the time difference will be too large.

3. Look at the pool status screen section on the dhcp leases page. All pools you defined (often 1 per interface) should be listed here. If any of the pools is in another state then "normal" you should debug the problem.

Author: anonymous via cvstrac, but he has some good items to shoot for. If the author happens to read this, please email me.