[ntp:questions] Re: NTP masters and slaves

Steve Kostecke kostecke at ntp.isc.org
Tue Nov 22 00:59:20 UTC 2005


On 2005-11-20, Jacob Friis Saxberg <webcom.dk at gmail.com> wrote:

> We have servers that we would like to be in sync via NTP. Below
> is the ntp.conf of our main server with the internal IP address
> 192.168.100.10.
>
> Could I just put "server 192.168.100.10" in the ntp.conf of all our
> clients to make them sync to the main server?

Yes. If you use iburst on your server lines you will reduce the intial
synchronization time from ~8 minutes to ~20 seconds. Your server lines
on the "client" should look like:

server 192.168.100.10 iburst

Don't forget to specify a drift file on the "clients".

Since you're having your clients depend on only one time server you need
to make sure that this server stays up and on time.

I suggest making the following changes to your main server ntp.conf:

> driftfile /etc/ntp/drift
> server  127.127.1.0
> fudge   127.127.1.0 stratum 10
> server pool.ntp.org

Increase the number of pool servers that you're using to 4, or more.
This will allow your "main server" to detect a bad time server. You may
be able to repeat the following line 4, or more, times:

server pool.ntp.org iburst

If you have resolver problems (i.e. you don't get different IP addresses
for each pool.ntp.org line, you may need to use some numbered sub-zones.
See the example at http://ntp.isc.org/Servers/NTPPoolServers

The following three lines aren't doing anything for you:

> multicastclient
> broadcastdelay  0.008
> authenticate no

-- 
Steve Kostecke <kostecke at ntp.isc.org>
NTP Public Services Project - http://ntp.isc.org/




More information about the questions mailing list