>>I use ntp-refclock. This way other machines on lan can syncronize to a
>>local ntp server, which in turn syncronizes with more precise machines.
>>In case of a network down when more precise machines are not reachable,
>>ntp-refclock can use local clock to provide ntp service to local
>>machines. This way you still have syncronized time on local lan.
>
>
> I think you've misunderstood what a "reference clock" is. It's an
> atomic clock such as is maintained at Stratum 1 sites. If you don't
> have one of those (*obviously* you don't) I think you chose the wrong
> one.
>
> I don't use Debian and don't know about the packaging, but my guess is
> that the -refclock package simply comes preconfigured to use a local
> reference clock, whereas the -simple one uses network Stratum 2
> servers.

English is not my first language so I may have not made myself clear
This is a snippet of my ntp.conf

server 127.127.1.0
fudge 127.127.1.0 stratum 5

Local server is set to a higher stratum, so if lower stratum servers are
available, local clock will never be used.

If for some reason lower stratum ntp server become unavaiable and local
clock synchronization is not set, local machines will not be able to
synchronize to local server and will start drifting apart.

When you do have local clock configured, local server will still provide
synchronization to local clients since it can reference local clock.
Therefore machines will not drift apart.

Also over the time, ntp server will figure out the drift of your clock
and even when lower stratum servers are not available, it will be able
to keep pretty precise time using this value. Atomic clocks is not the
only way to use refclock.