[Pdns-users] Virtual servers in pdns-recursor

[Miguel Miranda]

Well, after several hours of sending dnsdist stats i see embarrassing
numbers, considering the load and memory usage and still i got a lot 200 ms
response time (dnsdist.dns1j4s0n.main), what could be the cause of this,
network latency or am i totally messed up the recursor config? Each
instance has the same configuration:

setuid=pdns-recursor
setgid=pdns-recursor
daemon=yes
dont-query=127.0.0.0/8
local-address=127.0.0.1
local-port=5454
log-common-errors=no
max-cache-entries=1000000
max-negative-ttl=30
max-packetcache-entries=1000000
packetcache-servfail-ttl=30
quiet=yes
threads=4
trace=on






On Fri, Jan 8, 2016 at 4:26 PM, Miguel Miranda <miguel.mirandag at gmail.com>
wrote:

> 10 instances ??
> Im experimenting with 4 recursor instances because I have plenty of cpu
> and ram:
>
> top - 16:20:08 up 8 days,  7:36,  3 users,  load average: 0.06, 0.05, 0.05
> Tasks: 383 total,   1 running, 382 sleeping,   0 stopped,   0 zombie
> Cpu(s):  5.5%us,  2.5%sy,  0.0%ni, 91.4%id,  0.1%wa,  0.0%hi,  0.6%si,
>  0.0%st
> Mem:  32867864k total,  4847800k used, 28020064k free,   149300k buffers
> Swap: 16506872k total,        0k used, 16506872k free,  1395864k cached
>
>
> dnsdist is reporting around 1200 qps at peak hours, should i benefit with
> more simultaneous instances? if so, how many? Im starting to beleive this
> server is oversized for the dns traffic our customers are demanding, what
> do you think?
>
> On Fri, Jan 8, 2016 at 3:54 PM, Thomas Mieslinger <miesi at pc-h.de> wrote:
>
>> Am 08.01.2016 um 17:56 schrieb Nick Williams:
>>
>>> On Jan 8, 2016, at 9:46 AM, Pieter Lexis <pieter.lexis at powerdns.com>
>>> wrote:
>>>
>>>>
>>>> [..]
>>>>> /etc/init.d/pdns-recursor-balancer1 start.
>>>>>
>>>>
>>>> You'll need to add '--config-name'[1] to command line options
>>>> pdns_recursor in the init-script. Virtual hosting as done by the
>>>> authoritative server is not supported with the recursor init-script.
>>>>
>>>
>>> FWIW, this would be a good application for containers, such as Docker
>>> containers.
>>>
>>
>> I run 10 pdns_recoursor instances on each of my recursive server boxes.
>> I'm (still) using a distro with upstart so I just need a short
>> /etc/init/pdns-recursor.conf
>>
>> --snip--
>> description "pdns instance $INSTANCE"
>>
>> stop on runlevel [016]
>>
>> respawn
>>
>> instance $INSTANCE
>>
>> exec /usr/sbin/pdns_recursor --daemon=no
>> --config-dir=/etc/pdns-recursor/$INSTANCE --config-name=$INSTANCE
>> --socket-dir=/var/run/pdns-recursor/$INSTANCE
>> --snap--
>>
>> and can as many instances as I have CPU and ram. initctl list nicely
>> lists all running instances... restarts failed processes etc.
>>
>> Copying sysV initscripts around is not elegant, but with well written
>> upstart/systemd/svcadm configuration there is absolutely no reason to wrap
>> docker around pdns_recursor or pdns.
>>
>> Sorry
>>
>> Thomas
>>
>>
>>
>> _______________________________________________
>> Pdns-users mailing list
>> Pdns-users at mailman.powerdns.com
>> http://mailman.powerdns.com/mailman/listinfo/pdns-users
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.powerdns.com/pipermail/pdns-users/attachments/20160109/1817c8a7/attachment-0001.html>