[Pdns-users] High latency on recursion without cache

[Ciro Iriarte]

2015-01-19 16:57 GMT-03:00 bert hubert <bert.hubert at powerdns.com>:

> On Fri, Jan 16, 2015 at 01:34:45PM -0300, Ciro Iriarte wrote:
> > This were the queries sent from a client:
> >
> > 2015-01-16 13:04:38 dig @server 3.redhat.pool.ntp.org <-- NXDOMAIN, my
> > mistake
> > 2015-01-16 13:15:35 dig @server 2.redhat.pool.ntp.org <-- NXDOMAIN, my
> > mistake
> > 2015-01-16 13:15:52 dig @server 2.centos.pool.ntp.org <-- NOERROR
>
> Hi Ciro,
>
> Sadly your trace is interleaved with a bunch of other queries which make it
> harder to see what is going on.
>
> The trace for '2.centos.pool.nto.org' stops in the middle:
>
> Jan 16 13:15:53 rdns4m01 pdns_recursor[7294]: [2]    dns2.udel.edu.:
> accept answer 'adns3.upenn.edu.|AAAA|2607:f470:1003::3:c' from 'edu.'
> nameservers? YES!
> Jan 16 13:15:53 rdns4m01 pdns_recursor[7294]: [2]    dns2.udel.edu.:
> determining status after receiving this packet
> Jan 16 13:15:53 rdns4m01 pdns_recursor[7294]: [2]    dns2.udel.edu.: got
> NS record 'udel.edu.' -> 'dns1.udel.edu.'
>
> How did you gather or select the trace output? Is it possible you missed
> some output?
>
> Also, what Lua scripts do you run? Might these be interfering in some way?
>
>         Bert
>

Hi!, it seems that the rsyslog rate limit (default) feature is working too
well. I disabled that for this trace capture, I also disabled the Lua
script just in case, but the timing is roughly the same. This is a new try:


### Query from client
https://paste.ee/p/ov8EY

### Trace on the server
https://paste.ee/p/JLtOX

### Traffic dump
https://www.cloudshark.org/captures/9f9ee80f40fe

Regards!,

-- 
Ciro Iriarte
http://iriarte.it
--
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.powerdns.com/pipermail/pdns-users/attachments/20150120/f2423a78/attachment-0001.html>