[ntp:questions] Re: "Proper" way to test simple, new NTP configuration?

ohaya ohaya at cox.net
Fri May 20 20:01:59 UTC 2005


> If the system time is more than 1000 seconds off your NTP server's time then
> the NTP daemon terminates itself and prints a warning to the syslog that
> you should first set the clock manually.
> 
> On a client you can run
> 
>   ntpdate <server>
> 
> with <server> being the IP address of your NTP server, to get the client's
> time close to the NTP server's time fefore you start (x)ntpd on the client.
> 
> In recent NTP versions you can also start ntpd with the -g parameter, which
> has the same effect.
> 
> NEVER change the time manually on a machine which is synchronized by NTP. If
> you want to check that a client's time is synchronized, you can run
> 
>   ntpq -p hostname
> 
> (if you omit hostname, it defaults to localhost)
> which prints the time sync status on hostname, something similar to this:
> 
>      remote    refid    st t when poll reach   delay   offset  jitter
> =====================================================================
> *172.16.3.229  .PPS.     1 u   47   64  377    0.459    0.003   1.705
> 
> if there is a '*' at the beginning of one table entry then the NTP daemon
> has selected that time source as reference time source. The offset column
> shows the estimated time difference between the client and the server in
> milliseconds (!).
> 
> For an basic introduction to NTP, you may want to have a look at our NTP
> info page at
> http://www.meinberg.de/english/info/ntp.htm
> 
> or the support pages at http://ntp.isc.org.


Martin,

Thanks for the explanation, especially about ntpq.  I've now created a
small script that we can run which does an ntpq to all the servers, so
we can quickly check that all of them are synching to solaris1 (in my
case).

Jim



More information about the questions mailing list