[Pdns-users] PDNS Recursor on Debian / Low performance !!!URGENT!!!!

[Maroon Ibrahim]

Please check the stats of my server in production use.... always cache HITS 11% 

Feb  8 09:42:30 LebolDns pdns_recursor[1970]: Timeout from remote TCP client 109.75.64.121
Feb  8 09:42:31 LebolDns pdns_recursor[1970]: Timeout from remote TCP client 109.75.64.121
Feb  8 09:42:44 LebolDns pdns_recursor[1970]: Timeout from remote TCP client 109.75.64.121
Feb  8 09:42:45 LebolDns pdns_recursor[1970]: Timeout from remote TCP client 109.75.64.121
Feb  8 09:43:12 LebolDns pdns_recursor[1970]: stats: 8624720 questions, 918037 cache entries, 78819 negative entries, 11% cache hits
Feb  8 09:43:12 LebolDns pdns_recursor[1970]: stats: throttle map: 212, ns speeds: 3373
Feb  8 09:43:12 LebolDns pdns_recursor[1970]: stats: outpacket/query ratio 47%, 0% throttled, 0 no-delegation drops
Feb  8 09:43:12 LebolDns pdns_recursor[1970]: stats: 1933 outgoing tcp connections, 10 queries running, 84917 outgoing timeouts
Feb  8 09:43:12 LebolDns pdns_recursor[1970]: stats: 500083 packet cache entries, 58% packet cache hits
Feb  8 09:43:12 LebolDns pdns_recursor[1970]: stats: 51 qps (average over 1804 seconds)

I have 3Mbps full fiber dedicated for this server and i'm having DNS alerts on my bluecoat (cacheflow)....

Please advise 





> Date: Mon, 7 Feb 2011 11:37:46 +0100
> From: bert.hubert at netherlabs.nl
> To: maroon_ibrahim at hotmail.com
> CC: pdns-users at mailman.powerdns.com
> Subject: Re: [Pdns-users] PDNS Recursor on Debian / Low performance !!!URGENT!!!!
> 
> On Mon, Feb 07, 2011 at 09:38:42AM +0000, Maroon Ibrahim wrote:
> > all-outqueries  501740
> > outgoing-timeouts       352596
> > throttled-outqueries    76300
> 
> This mostly tells the story. Your pdns_recursor has problems talking to the
> network at sufficient speed. Either this is caused by you requiring too much
> bandwidth, or by a firewall that gets upset by all the different UDP
> 'sessions' it sees.
> 
> Of 501k queries, 428k did not get answers. 352k were direct timeouts, 76k
> were not even sent out by PowerDNS because it had a 'cached timeout'.
> 
> > over-capacity-drops     31945
> 
> You should increase 'max-mthreads' beyond 1024 after you have your network
> fixed, you dropped 31945 queries because 1024 were already running.
> 
> Fixing the network needs to happen upstream, this can't be done by PowerDNS.
> 
> Good luck!
> 
> 	Bert
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.powerdns.com/pipermail/pdns-users/attachments/20110208/e7d3608c/attachment-0001.html>