Question regarding names used for DDNS

Uwe Meyer-Gruhl dhcp_email at congenio.de
Mon Dec 3 16:12:11 UTC 2012


Nobody answered to my DDNS question yet, but I think I have found the 
answer in the meantime:

If one needs the ddns-hostname and ddns-domainname to overrule the FQDN 
(option 81) data coming from the client, the parameter setting "deny 
client-updates" is needed, like this:

host ddns-0000017 {
         option host-name "labprinter1";
         hardware ethernet 00:22:33:44:55:66;
         option domain-name "unknown";
         deny client-updates;
         ddns-updates on;
         do-forward-updates on;
         do-reverse-updates on;
         ddns-hostname "labprinter1";
         ddns-domainname "unknown";
}


Actually, I think this is a bug, since that parameter is only remotely 
connected to the issue in question. It should only keep the client from 
performing the DDNS updates by itself and not control if the server 
should accept the client's notion of its own name (and domain) when the 
server is responsible for the update.

This is probably a side-effect of the code changes for DDNS that have 
been applied in order to make DDNS updates asynchronous. I have found 
that the string definitions in server/dhcpd.c for std_nsupdate and 
old_nsupdate are now mostly irrelevant, since the former is evaluated 
only once and the latter cannot be used at all, because 
"ddns-update-style ad-hoc" leads to a configuration error in the first 
place in current versions.

Matter-of-fact, instead of seemingly configurable DDNS behavior, 
everything is hard-coded in server/ddns.c, the spaghetti code in which 
(lots of GOTOs!) has the side-effect noted above.

It also seems like these DDNS modifications brought another bug along: 
The SOA DNS server is not found automagically any more (i.e., the 
"primary server" part is always needed), which is a big problem in 
larger network installations.

So, to me it looks as DDNS support is treated as an orphan in ISC DHCP 
at this time (only emphasized by ISC DHCP's still ongoing disuse of the 
DHCID RR type, which has been around since 2006 already).



More information about the dhcp-users mailing list