Maintained by: NLnet Labs

[Unbound-users] No failover in stub-zone?

lst_hoe02 at kwsoft.de
Tue Jul 10 13:53:46 CEST 2012


Zitat von "W.C.A. Wijngaards" <wouter at nlnetlabs.nl>:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi Andreas,
>
> On 07/10/2012 01:02 PM, lst_hoe02 at kwsoft.de wrote:
>>
>> Zitat von "W.C.A. Wijngaards" <wouter at nlnetlabs.nl>:
>>
>>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>>
>>> Hi Andreas,
>>>
>>> On 07/09/2012 10:52 PM, lst_hoe02 at kwsoft.de wrote:
>>>> Hello
>>>>
>>>> we have on our border dns recursor (unbound 1.4.17) some
>>>> stub-zones mostly for PTR lookups for our internal AS112
>>>> addresses like this:
>>>>
>>>> stub-zone: name: "10.in-addr.arpa" stub-addr:
>>>> <IP-first-internal-NS> stub-addr: <IP-second-internal-NS>
>>>
>>> unbound will divide the load amongst the addresses. It will
>>> randomise with RTT banding.
>>>
>>>> Today the first internal NS went down and most reverse lookups
>>>> slow to crawl. I expexted unbound would notice the failure and
>>>> simply only use the second after some time like it did with
>>>> normal lookups when skipping unavailable NS.
>>>>
>>>> Is this expected behaviour or have i done something wrong?
>>>
>>> The second server also fails?
>>>
>>> Unbound should try both servers (randomly if they are working,
>>> for 50% load on both of them).
>>>
>>
>> No, the second was available, and yes it looks like Unbound was
>> balancing because some lookups where fast and some timeout. As far
>> as i know Unbound does skip unresponsive servers when doing
>> "normal" lookups (no stub-zones) and i suspected Unbound doing the
>> same for the stub-zone servers. Might this be possible as a feature
>> in the future? I think the same rules should apply for stub-zones
>> as for all lookups, no?
>
> This is the way it is implemented today.  Unbound can failover for
> stub-zones (and forward-zones) if nameservers do not respond and stops
> asking if they are down.
>

Strange, it looks like this has not happened. The primary was down for  
around half an hour already, but Unbound seems still trying to reach  
the stub-zone primary for part of the queries which in turn time out.
I will test as soon as possible and give more details on this.

Regards

Andreas