UCS 4.4-8 DNS-Client-Update/registerdns destroys DNS-record in some cases


i try to debug a weird occurence: In our DHCP DNS Settings policy dns client updates are disabled.

When (some) Clients request a DNS Update though, either automatically or via (ipconfig /registerdns) their DNS entry in Samba4-LDAP gets scrambled (DC=ts1,DC=our.domain.de,CN=MicrosoftDNS,CN=System,DC=our,DC=domain,DC=de).

What I mean with scrambled is that the dnsRecord HEX value changes and the flag dnsTombstoned=TRUE flag is set and bind does not resolve this host (ts1) anymore after this, instead bind answers with the catchall entry then.

This is what happens in the bind log:

Jun 16 13:28:30 dc1 named[48459]: samba_dlz: starting transaction on zone our.domain.de
Jun 16 13:28:30 dc1 named[48459]: client updating zone 'our.domain.de/NONE': update unsuccessful: ts1.our.domain.de/CNAME: 'rrset does not exist' prerequisite not satisfied (YXRRSET)
Jun 16 13:28:30 dc1 named[48459]: samba_dlz: cancelling transaction on zone our.domain.de

The only thing we can do is to delete the entry (and it’s reverse pendant) in OpenLDAP or the UCS DNS settings and readd it manually so it gets synced to Samba4-LDAP, then bind will resolve it again until the client tries to register this entry again. This happens for on windows Domain Member and for a Mac OS X Server.

Is our catchall entry an issue? Does anyone else has this problem? You help is much appreciated!

Thank you very much.