> Don't overestimate how accurate the monitoring system is.  There are lots of
> factors that can shift the accuracy by a few tens of ms from time to time.
> If your server is in agreement with various well run servers (nearby and
> farther away) then trust that. 
> 
> The monitoring system isn't intended to be accurate to less than 10ms; indeed
> a server doesn't get penalized until it's about 100ms off (!)[1].  Use the
> graphs for general trends and such, but don't believe it's always 100% right.
> It doesn't need to be to serve its purpose.

Ask,

What if the monitoring system Asks (sorry :) the server's
own estimation about its offset? (I mean value 1.780 below) (If possible.)
It depends less on _your_ network environment and probably
more realistic than a single poll from you.

ntpq> rl
associd=0 status=0615 leap_none, sync_ntp, 1 event, clock_sync,
version="ntpd [email protected] Sun Oct 17 13:45:13 UTC 2010 (1)",
processor="i686", system="Linux/2.6.32-5-686", leap=00, stratum=2,
precision=-21, rootdelay=0.271, rootdisp=33.834, refid=185.219.2.214,
reftime=d3c4be31.72b622fb  Thu, Aug  2 2012 10:52:01.448,
clock=d3c4c1ed.c0462e79  Thu, Aug  2 2012 11:07:57.751, peer=4972, tc=10,
mintc=3, offset=1.780, frequency=-304.690, sys_jitter=1.258,
clk_jitter=0.335, clk_wander=0.036, tai=35, leapsec=201207010000,
expire=201212280000
ntpq> 

I'd happy to enable your monitoring host to poll variables from
my servers.

Regards

Gabor
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool

Reply via email to