[Pdns-dev] statistics logging by pdns-recursor

Arjo Hooimeijer arjoslists at deathstar.xs4all.nl
Tue Jan 23 22:55:58 CET 2007


bert hubert wrote:
> On Tue, Jan 23, 2007 at 02:20:44PM +0100, Arjo Hooimeijer wrote:
>> I log all LOG_ERR prio messages to /dev/console which now fills up with
>> interesting but definitely non-crucial :-) information..
>> Is there any specific reason why pdns-recursor logs its statistics with
>> priority LOG_ERR?
> 
> Well, not really. Some time ago we shifted some of the logging priorities so
> they made more sense, this wasn't noticed. LOG_WARNING is quite often not
> logged at all, that might've been the reason.
> 
> The whole syslog severity levels thing is a tad odd, 'LOG_ERROR' is in fact
> in the middle of all severities, with LOG_WARNING often turning out not to
> be logged at all.
> 
> If we look at the scope of things, a running powerdns recursor will log
> quite some information at LOG_ERROR already.
> 
> So I'm not inclined to really change this, as it won't change the amount of
> data logged for most people in a real sense, unless your PowerDNS is very
> quiet.
> 
> But I've been known to change my mind :-)
> 
I did a grep on the sources and this shows me that LOG_CRIT is used for
the real issues, so I can easily just tweak my syslog.conf. I don't have
a problem with the curious setup, just being curious about the backgrounds.

It appears to me that this level of verbosity would justify LOG_INFO
rather than LOG_ERR, which should, in my opinion, be used for messages like:
pdns_recursor.cc:    L<<Logger::Error<<"startDoResolve problem:
"<<ae.reason<<endl;
pdns_recursor.cc:    L<<Logger::Error<<"DNS parser error:
"<<dc->d_mdp.d_qname<<", "<<e.what()<<endl;
pdns_recursor.cc:    L<<Logger::Error<<"STL error: "<<e.what()<<endl;

Of course, you are right the amount of sysadmins that choose to simply
ignore everything below LOG_ERR level, but then again, would they be
interested in recursor statistics? ;-)


More information about the Pdns-dev mailing list