Hi Aki, Jacky,
This is explained in
http://doc.powerdns.com/html/recursor-design-and-engineering.html#idp10119168 :
"Whenever a packet is sent to a remote nameserver, the response time is stored
in the SyncRes::s_nsSpeeds map, using an exponentially weighted moving average.
This EWMA averages out different response times, and also makes them decrease
over time. This means that a nameserver that hasn't been queried recently
gradually becomes faster in the eyes of PowerDNS, giving it a chance again."
Bert
On 27 Dec 2013, at 08:50, Aki Tuomi <cmo...@youzen.ext.b2.fi> wrote:
> once cache expires the records are removed and are looked for again. if you
> look
> at SyncRes::getAddrs, you'll see the record selection process.
>
> btw, are you just interested or is there some deeper issue here?
>
> Aki
>
>
> On Fri, Dec 27, 2013 at 03:29:27PM +0800, jacky wrote:
>> hi, Aki tnuomi，
>> Thank you for suggestion. I found pdns always get the fastest one from the
>> front of result returned by function "SyncRes::shuffleInSpeedOrder”. If a
>> slow one at the back has no opportunity to be choosed, how pdns know it
>> become fast. I can not find the source code make pdns query the slow one.
>>
>> on the other hand, I found there is a houseKeeping thread which will erase
>> the stale speed value of a dns server . If a slow dns at the back of
>> t_sstorage->nsSpeeds map, it will be erased because of rarely to be queried,
>> and only fast dns server will have its speed updated by this code "
>> t_sstorage->nsSpeeds[*tns].submit(*remoteIP, lwr.d_usec, &d_now);” at
>> syncres.cc:1003.
>> . And then, only dns servers which are mostly queried will stay in
>> t_sstorage->nsSpeeds map. It make me puzzled.
>>
>> --
>> jacky
>>
>>
>> On Friday, December 27, 2013 at 3:08 PM, Aki Tuomi wrote:
>>
>>> On Thu, Dec 26, 2013 at 10:34:23PM +0800, jacky wrote:
>>>> hi, all,
>>>> I think may be it is a question more suitable to dev maillist, so I submit
>>>> it here. If I made a mistake , please let me know. Thanks.
>>>>
>>>> I'am reading the source code of pdns-recursor-3.5.3. There is question I
>>>> can not understand.
>>>> When a authoritative dns server become slow because of overload,
>>>> pdns-recursor will move it backwards in the
>>>> t_sstorage->nsSpeeds map, and then a faster dns server will be used
>>>> preferred. But when the dns server recoverd
>>>> from overload and become fast again, what will pdns-recursor do about it ?
>>>> Will pdns-recursor move it forward
>>>> in speed map? I can't find relevant source code. Any suggestion?
>>>>
>>>> --
>>>> jacky
>>>>
>>>
>>>
>>> Hi!
>>>
>>> It works by moving the fastest one on top.
>>>
>>>> _______________________________________________
>>>> Pdns-dev mailing list
>>>> Pdns-dev@mailman.powerdns.com (mailto:Pdns-dev@mailman.powerdns.com)
>>>> http://mailman.powerdns.com/mailman/listinfo/pdns-dev
>>>>
>>>
>>>
>>>
>>
>>
> _______________________________________________
> Pdns-dev mailing list
> Pdns-dev@mailman.powerdns.com
> http://mailman.powerdns.com/mailman/listinfo/pdns-dev
_______________________________________________
Pdns-dev mailing list
Pdns-dev@mailman.powerdns.com
http://mailman.powerdns.com/mailman/listinfo/pdns-dev
