[Pdns-users] Spikey response times in powerdns recursor

[bert hubert]

On Wed, Mar 17, 2010 at 11:37:24AM +0000, Simon Bedford wrote:
> >bbc.co.uk still has a nameserver that is down, so having that domain resolve
> >slowly every once in a while is to be expected.
> 
> Agreed, there is the point that this does not happen in the day
> though and although the cache will be busier it will still have to
> go get the short lived TTL entry every x minutes in the day as well
> unless I am missing something.

Let's say you have 100 customer originated lookups per 5 minutes during the
day for bbc.co.uk, and only 10 per 5 minutes at night.  This means that your
script has ten times less chance of hitting the cache miss during the day.
Hence the fact that you observe the problem with this problem domain more.

> >You've indicated you've occasionally seen 500ms lookups times for
> >google.com, but I have not heard of any other problems.
> >
> >google.com takes between 0 and 100ms to resolve in my tests.
> 
> hmmm, we definitely see higher as we do with our own domains
> overnight as well.

Please show me data of all these domains. It is very hard to solve problems
withing having real numbers and real details.

> I realise dnsscope takes all domains into account and agree that the
> average stats are pretty darn good.  But we do see multiple second
> look up times from the recursor for domains we host and run the auth
> DNS for without seeing those spikes on our auth DNS graphing.

That is probably due to 3.1.x. 

> This is what is causing the mystery for me, when its good its really
> good but then response times go crazy at a random time, its dropped
> our customer experience graphing from 99.987% to 89% (some of this
> will be the 3.1.7.2 cache maintenance bug though, in fact a larger
> proportion as we only have 1 of 4 upgraded to 3.2.1).

I suggest upgrading the rest too, since it will definitely get rid of the
multi-second delays.

Regarding the google lookups, try: 
$ dig www.google.com @216.239.38.10

This is ns4.google.com, and it reliably takes 97ms to answer a query from
here (The Netherlands).

So it is not surprising that every once in a while, a lookup for
www.google.com takes 97ms.

	Bert