AXFR resolution path

Kevin Darcy kcd at daimlerchrysler.com
Mon Sep 26 23:09:22 UTC 2005


Jeffrey Stevens wrote:

>If I configure my DNS to be master for a fake test zone and attempt to use dig
>to that server to transfer the zone, it times out.  I believe it is recursing to
>the root servers to resolve the (internal intranet) MNAME in the SOA.
>
>Ok, I thought I'll first have it resolve the MNAME (should get it from a
>forwarder first, not a root) and then try the zone transfer.  This should avoid
>the recursion to the roots, but does not, it just times out.  Did I understand
>this correctly?
>  
>
Zone transfers are not recursable. If you're pointing your client 
directly at the master or one of the slave servers (e.g. via dig 
example.com axfr @x.x.x.x) and it's timing out, then there some sort of 
connectivity and/or firewall-rule problem involved. Firewall-rule 
problems aren't terribly unusual, since zone transfers use TCP, whereas 
most (but not all) ordinary DNS queries use UDP -- certain overzealous 
and/or ignorant firewall admins will allow the "safe" UDP protocol 
through the firewall, but block the "scary" TCP protocol.

Not sure where you're going with the whole MNAME thing. For zone 
transfers, you need to talk directly to an authoritative nameserver for 
the zone. The MNAME _should_ refer to an authoritative nameserver for 
the zone, but many folks screw up their MNAME fields with little or no 
ill effect (i.e. there's little incentive for them to fix it), and even 
if the MNAME field is "correct", it might refer to a "hidden master" 
that you can't even talk to. You'd be better off stepping through the NS 
records for the zone, but ultimately, if this is a zone you have any 
legitimate business zone-transferring, you should already be in contact 
with the administrator, and they'll tell you what names and/or IP 
addresses to use. You might also need to configure the appropriate TSIG 
keys, if transfers of the zone are crypto-authenticated.

                                                                         
                                                                     - Kevin




More information about the bind-users mailing list