At the company I used to work for, we did this sort of syncronization
with ntp on numerous isolated networks withing manufactoring plants.
Unfortunately I do not have those configuration files with me, but they
worked something like this. We would pick one or two machines to be the
"masters". On those, ntp.conf would have a server line pointing to the
internal clock. The ntp.conf would also have a peer line pointing to the
other master. The ntp.conf in the other machines would have server lines
pointing to these masters. It took a little fiddling to get the
configurations correct, but once we did, ntp worked amazingly well at
keeping all the clocks syncronized, even though they would drift from
the "One True UTC".
Hope this helps.
Andrew Robinson
Brian T. Brunner wrote:
I have checked into ntp/ntpd and found it to be broken too far for
mere words.
It's avowed purpose is to find the One True UTC, I have an
isolated gaggle of
machines on a private net that need to stay together regardless of how
far that
'together' drifts from WWV or some net-readable atomic clock. Given a
gaggle
with no net connect to other machines outside the gaggle, ntp fails to
perform,
and fails to explain why.
ntp's documentation frustrates and obfuscates. Unless there is an
ntp HOWTO
for dummies, I've dumped ntp.
--
Shrike-list mailing list
Shrike-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/shrike-list
