Failover : one DHCP cannot delete TXT/PTR records (timed out)

[Nicolas C.]

Hello,

I'm using DHCP in failover mode, one server hosts the DHCP and the DNS, 
the other only DHCP, dynamic DNS updates are activated ("deny 
client-updates").

Our version of DHCP is the one provided by Debain Squeeze : 4.1.1-P1.

The DHCP configuration is the same on the two servers (updated with 
rsync) at the exception of the "failover peer" section witch is unique 
on each server.

I recently noticed that the "dhcp-only" server cannot cleanup the DNS 
correctly, it can delete the "A" record but the PTR and TXT are left behind.

I don't understand what's happening. Somebody had a similar issue : 
https://lists.isc.org/pipermail/dhcp-users/2012-February/014965.html

Below is a overview of the problem, I'm not sure if it is relevant but 
bind is sending notifies before the TXT/PTR deletion attempt. We are 
using IPv4 and IPv6 but on the Bind log I can see the "dhcp-only" server 
using its IPv4 or IPv6 address with no problem.

Any help or ideas on this issue would be greatly appreciated.

=> DHCP log : delete of "A" record, OK

Jul 25 21:25:58 tunng dhcpd: if VIRTUP1-VB5S480.mydomain.ext IN TXT 
"31dcd9480a551fe071cb967d1eb2b2e0db" rrset exists and 
VIRTUP1-VB5S480.mydomain.ext IN A 172.20.24.98 rrset exists delete 
VIRTUP1-VB5S480.mydomain.ext IN A 172.20.24.98: success.


=> DHCP log (continuing) : delete of "TXT" record, not OK

Jul 25 21:25:59 tunng dhcpd: if VIRTUP1-VB5S480.mydomain.ext IN A rrset 
doesn't exist and VIRTUP1-VB5S480.mydomain.ext IN AAAA rrset doesn't 
exist delete VIRTUP1-VB5S480.mydomain.ext IN TXT 
"31dcd9480a551fe071cb967d1eb2b2e0db": timed out.

=> Bind log : only the "A" deletion is visible

Jul 25 21:25:58 calla named[1149]: client 172.24.0.2#49734: updating 
zone 'mydomain.ext/IN': deleting an RR at VIRTUP1-VB5S480.mydomain.ext A
Jul 25 21:25:58 calla named[1149]: zone mydomain.ext/IN: sending 
notifies (serial 2010367927)

Regards,

Nicolas