[Pdns-users] PowerDNS Recursor 3.2 Release Candidate 1 available
bert.hubert at netherlabs.nl
Fri Feb 19 10:29:57 UTC 2010
Thanks for your report!
On Fri, Feb 19, 2010 at 11:22:31AM +0100, Peter Gervai wrote:
> But here's what made me write: cache efficiency. Old version forks ran
> at average 70% (both). New version right now runs at 25%. Since
> nothing was really changed (userbase same, config same, cache size is
> halved [3+3 down to 3] but in the past I've observed no real reduction
> in cache efficiency between 1M [=2M] and 3M [=6M] entries so I
> wouldn't guess this is the reason), I really wonder why.
I wrote this up yesterday,
The final line explains this:
'The Packet Cache is consulted first, immediately after receiving a packet.
This means that a high hitrate for the Packet Cache automatically lowers the
cache hitrate of subsequent caches. This explains why releases 3.2 and
beyond see dramatically lower DNS cache hitrates, since this is the first
version with a Packet Cache.'
> Then, the sum of <answers* > doesn't add up to anything I see (like
> <noerror-answers>), like they used to before. It's more like the half
> of the queries, which is odd.
And this is explained on:
http://doc.powerdns.com/recursor-stats.html (updated yesterday):
'It should be noted that answers0-1 + answers1-10 + answers10-100 +
answers100+1000 + packetcache-hits + over-capacity-drops = questions.'
The main thing is to add the 'packetcache-hits'.
These were the 'DOCUMENTATION FORTHCOMING' notices in the release notes that
have now been replaced by content.
It appears the documentation was updated in exactly the right places ;-)
More information about the Pdns-users