I find that xntp does not adjust the realtime hardware BIOS clock, only the
system clock. Over time, if a significant skew builds up, on the next
reboot the clock will be thrown back.
If the master is rebooted, when there's a skew, all the slaves will get
their clocks reset. If the slave is rebooted, several minutes will go by,
then the slave's clock will be jump by a large amount, when it is
resynchronized with a master.
Solution: add the following to /etc/rc.d/init.d/xntpd, after the killproc
that stops the daemon:
This will adjust the hardware clock with the accumulated skew, before the
xntpd should update the RTC itself. Are you running on Compaq ProLiant hardware?
I think i noticed this problem on ProLiants (hwclock --systohc being unable to
update the RTC).
*** This bug has been marked as a duplicate of 1142 ***