DHCP upgrade = crash

[Nicholas F Miller]

We upgraded from dhcp-3.1.1 to dhcp-4.1.0p1 last week. About a hour  
after the upgrade DHCP seemed to freeze up. If you ran tail -f on the  
dhcp.leases file you would see leases continuing to be handed out but  
it wasn't logging to the daemon log. We reverted back to dhcp-3.1.1  
after a second freeze up and thing worked fine.

Looking at the logs it seems that a machine on our network was doing  
around 23 DHCPDISCOVERs a second. In one day the daemon log contained  
36,706,541 entries, 35,606,198 of which were for this one machine. The  
CPU load on the server while this was going on, with either DHCP  
version, was only around 5-6%. One other interesting thing I found was  
the last thing in the daemon log before both freezes was a DHCPNAK.

In case it matters, when I compiled dhcp-4.1.0p1 I used the --prefix,  
--localstatedir and --disable-dhcpv6 flags.

Why is dhcp-3.1.1 able to keep up with this amount of traffic but  
dhcp-4.1.0p1 is not?
_________________________________________________________
Nicholas Miller, ITS, University of Colorado at Boulder