[Unbound-users] Forward zone query distribution

[İhsan Doğan]

Hi,

I've got a question how queries are distributed to forward
resolvers.

In this example, I've defined that all queries, which can't be
answered from the cache, are forwarded to Google's DNS servers:

forwared-zone:
name: "."
forward-addr: 8.8.8.8
forward-addr: 8.8.4.4

How are the queries distributed between 8.8.8.8 and 8.8.4.4? Are
all queries going to 8.8.8.8 because it's the first one, are they
distributed via round-robin or is the one with the lowest
response time getting the most of the queries?




Ihsan

[W.C.A. Wijngaards]

Hi İhsan,

Post by İhsan Doğan
Hi,
I've got a question how queries are distributed to forward
resolvers.
In this example, I've defined that all queries, which can't be
8.8.4.4
How are the queries distributed between 8.8.8.8 and 8.8.4.4? Are
all queries going to 8.8.8.8 because it's the first one, are they
distributed via round-robin or is the one with the lowest response
time getting the most of the queries?

Initially at random. Then, once the response time is known, one of
the fastest ones. It picks one at random from the "RTT band", so
small differences in response time do not matter. If ping time to
those two IP addresses is roughly similar, then it'll end up as a
50/50 split.

There is a lot more logic, mostly for interoperability with
malconformant servers or servers with DNSSEC-bogus content. But for
this example that would not make a difference.

Best regards,
Wouter

Post by İhsan Doğan
Ihsan

[İhsan Doğan]

Hi Wouter,

Post by W.C.A. Wijngaards

Post by İhsan Doğan
I've got a question how queries are distributed to forward
resolvers.
In this example, I've defined that all queries, which can't be
8.8.4.4
How are the queries distributed between 8.8.8.8 and 8.8.4.4? Are
all queries going to 8.8.8.8 because it's the first one, are they
distributed via round-robin or is the one with the lowest response
time getting the most of the queries?

Initially at random. Then, once the response time is known, one of
the fastest ones. It picks one at random from the "RTT band", so
small differences in response time do not matter. If ping time to
those two IP addresses is roughly similar, then it'll end up as a
50/50 split.
There is a lot more logic, mostly for interoperability with
malconformant servers or servers with DNSSEC-bogus content. But for
this example that would not make a difference.

Thanks a lot for the quick reply. :-)



Ihsan