[ntp:questions] Score is drop in cycles

Brian Inglis Brian.Inglis at SystematicSw.ab.ca
Fri Feb 14 20:32:25 UTC 2014


On 2014-02-14 08:07, mingdawang at 163.com wrote:
> Hi  All,
>
> 	I try add my ntp server to pool, but I found it's score is drop in cycles ,
>  I think it's related to traffic,but the system's load is very low, I didn't find
>any error in log. How to find the bottleneck of the system?
> Looking forward to your reply.
>
> 	http://www.pool.ntp.org/scores/114.113.198.166
>
> [root at hz1-12-168 ~]# uname -a
> Linux hz1-12-168.inter.163.com 2.6.32-431.5.1.el6.x86_64 #1 SMP  
> Fri Jan 10 14:46:43 EST 2014 x86_64 x86_64 x86_64 GNU/Linux
>
> [root at hz1-12-168 ~]# ntpd -!
> ntpd 4.2.6p5
> ntpd 4.2.6p5 at 1.2349-o Mon Jul 15 09:22:49 UTC 2013 (1)
>
> [root at hz1-12-168 ~]# ntpq -pn
>       remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
> -64.113.32.5     .ACTS.           1 u   27 1024  377  358.521  -30.035   6.725
> *137.189.4.10    .GPS.            1 u  750 1024  377   81.095   -3.238   8.187
> +108.71.253.18   .GPS.            1 u  374 1024  377  311.508   -2.693   5.882
> +83.163.209.244  .GPS.            1 u  416 1024  377  404.777   -1.397   7.461
> -89.109.251.23   .PPS.            1 u  486 1024  377  295.556  -37.660  50.758
> -193.190.230.65  .PPS.            1 u  546 1024  377  407.231  -17.679  19.307

Most of your delays are really long > 100ms which can make it difficult to keep
consistent time, with your offsets jumping between single and double digit ms,
and with monitoring dropouts, shown in the scores CSV log, reducing your points
by -5 to below 10 on your scores page every 3-4 hours.

You may be unable to do much about the dropouts with monitoring from Los Angeles,
but if you can make your time more consistent, you will have a higher score which
monitoring dropouts may not affect as much.

> [wangmd at hz1-12-168 ~]$ cat /etc/ntp.conf
>
> driftfile /var/lib/ntp/drift
>
> restrict default kod nomodify notrap nopeer noquery
> restrict -6 default kod nomodify notrap nopeer noquery

add restrict limited to kod to rate limit requests:
restrict    default kod limited nomodify notrap nopeer noquery
restrict -6 default kod limited nomodify notrap nopeer noquery

> restrict 127.0.0.1
> restrict -6 ::1
>
> server nist.netservicesgroup.com
> server clock.cuhk.edu.hk
> server t1.timegps.net
> server ntp.remco.org
> server ntp3.vniiftri.ru
> server ntp1.oma.be

You may get better results by dropping all servers except ...hk
and replacing them by pool servers (add iburst for faster startup):
server clock.cuhk.edu.hk iburst
pool     cn.pool.ntp.org iburst
pool   asia.pool.ntp.org iburst

This may not be usable as a pool member, but you may be able to get in
contact via the pool to peer with more local servers having lower delay.

> includefile /etc/ntp/crypto/pw
>
> keys /etc/ntp/keys
>
> statistics clockstats cryptostats loopstats peerstats sysstats protostats

You may need to add before the line above:
enable stats

and you do not need clockstats if you are not running a local reference clock.

> 2014-02-14
> Mingda Wang
-- 
Take care. Thanks, Brian Inglis


More information about the questions mailing list