PowerDNS Recursor FAQ

Here you can find the answer to some of the most frequently asked questions about the recursor package within PowerDNS. If you have a question not answered by this page you can add

.. contents::
.. sectnum::

What kind of statistics are gathered by the powerdns_recursor?
==============================================================

::

  questions          counts all End-user initiated queries with the RD bit set
  nxdomain-answers   counts the number of times it answered NXDOMAIN since starting
  noerror-answers    counts the number of times it answered NOERROR since starting
  servfail-answers   counts the number of times it answered SERVFAIL since starting
  outgoing-timeouts  counts the number of timeouts on outgoing UDP queries since starting
  all-outqueries     counts the number of outgoing UDP queries since starting
  throttled-out      counts the number of throttled outgoing UDP queries since starting
  answers0-1         counts the number of queries answered within 1 milisecond
  answers1-10        counts the number of queries answered within 10 miliseconds
  answers10-100      counts the number of queries answered within 100 miliseconds
  answers100-1000    counts the number of queries answered within 1 second
  answers-slow       counts the number of queries answered after 1 second
  tcp-questions      counts all incoming TCP queries (since starting)
  tcp-outqueries     counts the number of outgoing TCP queries since starting
  nsspeeds-entries   shows the number of entries in the NS speeds map
  negcache-entries   shows the number of entries in the Negative answer cache
  throttle-entries   shows the number of entries in the throttle map
  cache-entries      shows the number of entries in the cache
  cache-hits         counts the number of cache hits since starting
  cache-misses       counts the number of cache misses since starting
  qa-latency         shows the current latency average
  concurrent-queries shows the number of MThreads currently running



The recursor tarball comes with a /rrd directory that includes ready-to-go scripts for generating graphs such as these from the statistics the recursor provides:

http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnsquestionsd790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnslatenciesd790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnshitrated790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnscachesd790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnsqoutqd790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnstimeoutsd790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnstcpquestionsd790-1.png
http://wiki.powerdns.com/cgi-bin/trac.fcgi/attachment/wiki/RecursorFAQ/pdnsloadd790-1.png


What's in the NS Speeds map?
============================
It is the amount of entries in the speed map. That map contains all the speeds and availability of the authoritative nameservers the recursor is talking with.

What's in the Throttle map?
===========================
This map contains a history about nameservers returning unuseful answers. The ''throttle-entries'' in the statistics shows the number of entries in this map.

What's the ''throttled-out'' in the stats?
==========================================
If a certain nameserver is not responding, or if it is, but returning unuseful answers, the query can
get prevented from happening and causing delay. This is called a 'throttled out-query'. Clients will receive a SERVFAIL answer until a trustworthy answer is found. When you see the throttled-out peaking abnormally high you can probably observe the same peak on the SERVFAIL answers.

I'm seeing abnormal high SERVFAILS on my backup resolver
========================================================
Backup resolvers typically see a higher than average of ''slow queries''. The answers for these queries are hard to find, and thus your customers software is shouting out more questions to more servers to get an answer.

My cache hit% is quite low
==========================
I (kai) have seen cache miss rates of around 20%, with spikes to 40% on a busy resolver (500 queries per second). A less busy cache shows a missrate of 30% with spikes to 70% (less than 300 queries per second).

Example graph for cache hits: [[Image(cache-hits.png)]]