On Wed, 2014-07-09 at 23:06 +0200, Jakub Hrozek wrote:
one of our users ran into an interesting problem -- her AD
infrastructure was different from the DNS server. Because by default, we
perform update against the server we're connected to, the DNS update
Per Simo's suggestion, I've implemented a new option that allows the
administrator to override the DNS server used for DNS updates.
I am not sure I like the premises here.
The problem is that we are trying to set a DNS server at all in the
The normal case should be that nsupdate is not passed any server name
and it discovers the correct server via normal DNS lookup queries.
If someone is using a messed up DNS setup then, yeah I would offer a
setting for specifying exactly what DNS server to contact, but I am not
sure it makes sense to fallback to a random Domain Controller by
IIRC the cldap ping reply does return you information on whether the
domain controller is also a DNS server, I think we need to extract that
info and keep it around, and fallback to directly contact the domain
controller only if the cldap ping indicated DNS support.
Comments on code follow:
I think I would change the following:
+ Setting this option makes sense for environments
+ where the DNS server is different from the identity
I think you should say something like this:
"Set this option if your DNS environment is broken and the correct
server cannot be discovered via regular DNS lookups."
On the code side I think you are missing to remove the code in
ad_dyndns.c:ad_dyndns_update_send() that unconditionally sets
stat->servername, by default no server name should be set.
Exactly the same issue is present in
ipa_dyndns.c:ipa_dyndns_update_send(), once again the current freeipa
server name is taken as a DNS server. And once again not all IPA servers
are DNS servers. Here too we need to remove all the code that sets the
server name and let nsupdate find the server through the normal dns
lookup queries or use the dyndns_server fallback if set.
Using unconditioanlly a domain controller you are randomly bound to is
normally more wrong than right and we should stop doing that completely.
So NACK, for now.
Simo Sorce * Red Hat, Inc * New York