[ntp:questions] Should ntpd log failure to syslog?

Uwe Klein uwe_klein_habertwedt at t-online.de
Wed Dec 10 18:31:51 UTC 2008


Richard B. Gilbert wrote:
> Uwe Klein wrote:
> 
>> Richard B. Gilbert wrote:
>>
>>> Normally, reaching a server is not a problem.  If ntpd can't reach 
>>> its configured servers, something is horribly wrong somewhere and 
>>> your network people should be working like beavers to fix it!
>>
>>
>> How should they know if ntp is mum about that failure?
>>
>> uwe
> 
> 
> IF NTP IS THE ONLY THING FAILING, they won't know.  It's probably not 
> their problem.  If you have four servers configured and you can't reach 
> any of them, it's almost certainly a network problem and your networking 
> people should know about it!
> 
> If you MUST have NTP working at all times, you need to build a defense 
> in depth.  That means one or more hardware reference clocks and/or four, 
> five, or seven upstream servers.  I use a GPS timing receiver as the 
> source for one of my servers, a WWV/WWVH radio clock for another, and 
> several internet servers as backup/sanity check.
> 
> And I'm just a hobbyist these days!  But I'm a hobbyist who knows what 
> time it is! :-)

So, you are a hobby rocket scientist?  ;-)

back to the topic:
with firewalls, overintelligent switches, NAT and nutty admins
it is no longer a given that reaching a host via ping, traceroute
or http means it works globally and you can now savor your next cigar.

I prefer information push of failures ( like to syslog )
to information pull ( like in writing utilities that monitor
uncooperative apps and which I have to setup as a cronjob ..... )

syslog already is a very flexible tool for monitoring and problem escalation.


uwe




More information about the questions mailing list