[Pdns-users] Glue-Records in PowerDNS 3.x

[Peter van Dijk]

Hello Marten,

On Dec 13, 2012, at 1:35 , Marten Lehmann wrote:

> Hello,
> 
> we recently upgraded from PowerDNS 2.9.21 to 3.1 and noticed a different handling of subdomains within subdomains, that contain NS-records.

There have been changes in that area, indeed - especially when running with -dnssec in pdns.conf.

> Let me show an example for the zone domain.com, which has an entry in the domains table (its a NATIVE zone) and its records in the records table related by the domain_id.
> 
> domain.com A 1.2.3.4
> www.domain.com A 2.3.4.5
> sub.domain.com NS whatever.com
> www.sub.domain.com A 3.4.5.6
> 
> domain.com and www.domain.com resolve as expected in both releases. When queried for www.sub.domain.com, in release 2.9.21 PowerDNS responds with 3.4.5.6 . In version 3.1 it doesn't find a A record.

Yes, this is correct. 2.9.21 would return the A, but 3.1 should return the delegation instead (at least, if -dnssec is set).

> Interestingly, in 3.1 it responds with a SOA record for sub.domain.com, meaning that by just having inserted an NS-record for a sub.domain.com, PowerDNS creates a virtual SOA record (there is no SOA-record in the records table for sub.domain.com) and doesn't look for www.sub.domain.com in the same zone any longer. It probably fails because it tries to lookup sub.domain.com again in the domains table but doesn't find a record (we didn't create a separate zone for it) so it doesn't resolve at all instead of using the record within the domain.com zone.

3.1 should not synthesize SOA records. Can you show actual output? We prefer it if you use real domain names, by the way. Feel free to insert domain.com into your database/backend if you must - just don't edit the output.

Actual backend data would also be useful to see.

Kind regards,
-- 
Peter van Dijk
Netherlabs Computer Consulting BV - http://www.netherlabs.nl/