BIND caching of nxdomain responses

Dan Hanks danhanks at gmail.com
Fri Oct 22 14:56:30 UTC 2021


On Fri, Oct 22, 2021 at 10:29 AM Matus UHLAR - fantomas
<uhlar at fantomas.sk> wrote:
>
> On 22.10.21 09:57, Dan Hanks wrote:
> >As I understand RFC 2308, when receiving an NXDOMAIN response, and when
> >deciding how long to cache that NXDOMAIN response, a resolver should use
> >whichever value is lower of the SOA TTL, and the SOA.minimum value as the
> >length of time to cache the NXDOMAIN.
> >
> >I have a situation where I am seeing different behavior from that in BIND.
> >Given the following SOA record:
> >
> >azure.mongodb.net.      900     IN      SOA     ns-1430.awsdns-50.org.
> >awsdns-hostmaster.amazon.com. 1 7200 900 1209600 60
> >
> >I am finding that BIND (9.11.x) is caching the NXDOMAIN response for 900s
> >(SOA TTL), instead of the expected 60s (SOA.minimum).
> >
> >I have noticed that many auth servers out there will drop the SOA TTL to
> >match the SOA.minimum value when attaching the SOA record to an NXDOMAIN
> >response. Is BIND expecting this to happen, and just opting to use the SOA
> >TTL value (and not the SOA.minimum value if they disagree)?
>
> are you authoritative server for azure.mongodb.net?
> if not, BIND will use cache time that came from authoritative server adn
> won't parse the SOA itself.

I am not authoritative, I'm just making recursive queries against this domain.

When you say, "BIND will use cache time that came from authoritative
server", what 'cache time' are you referring to? Are you referring to
the values in the SOA record included in the AUTHORITY section of the
NXDOMAIN response?

Thanks,

Dan


More information about the bind-users mailing list