Reply to remote unicast DHCP requests

Paúl Ortiz Imedio portizimedio at
Tue May 3 13:57:51 UTC 2011

>Paúl Ortiz Imedio wrote:
>> I am trying to figure out the behaviour of the DHCP server when replying to
>> unicast DHCP Requests from clients at remote subnets (RENEWING state).
Simon Hobson wrote:
>It's quite simple. The server should reply by unicast to the client. The 
>packet is routed via normal OS routing rules.
I agree. That behaviour is what I was expecting.
>When you say the server is not responding - does it log the request ? Does it 
>log that it's replying ? Does it have a route to the client's network ?
Server logs the request, but it does not answer:
DHCPREQUEST for from XX:YY:ZZ:XX:YY:ZZ via eth0: ignored (not authoritative)
There is no routing problem and it receives the request from the client. The
problem is that the server interface where it receives the request is not
attached to any shared network. The server finds the valid lease and the 
configured subnet for ciaddr, but the lack of giaddr makes it ignore the 
request. If I configured the server to be authoritative, it would send a 
DHCPNAK (that is not a solution). That is why I started looking at the code 
for an explanation and I found what I have detailed in my first post.
>For details of how the protocol works, I recommend "The DHCP Handbook"  by 
>Ralph Droms and Ted Lemon. There's a second edition which covers failover, 
>but for the basics the original edition will do fine if you can't find the 
>newer one.
I had already checked it in that book, as it has been recommended too in 
previous posts. And there I could confirm the expected behaviour.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the dhcp-users mailing list