Friday, September 7, 2007, 9:16:08 PM, you wrote:
On Fri, 2007-09-07 at 17:44 +0100, Andy Green wrote:
> ntp has a "sanity check" limit, if the actual clock and the ntp time
> differ by more than this (1000s? one thousand somethings IIRC) then it
> just gives up.
Ok, I stopped ntpd, ran ntpdate manually and it synced the clocks.
Then I started ntpd again and it still couldn't get synced. So it's
not a 'sanity check' who is not letting the ntpd work.
Make sure you have the IP addresses of your NTP servers in BOTH the
/etc/ntp/ntpservers and /etc/ntp/step-tickers files. The first thing
the startup script does is try to run "ntpdate" using the entries in
the ntpservers file. That will drag your clock into sync with the
servers by brute force, then ntpd can keep it synced.
Ok, I checked the /etc/ntp/ntpservers and they are the same on both
Fedora Core 6 and Fedora 7 machines containing just 2 lines:
clock.redhat.com
clock2.redhat.com
/etc/ntp/step-tickers are empty on both machines. I have put the same
lines there from 'ntpservers' file but it didn't help.
Oh! I forgot to mention one thing. ntpdate by itself doesn't
work without any parameters given. I have to manually point it to an
ntp server - only then the time gets synced:
[root@frontend ~]# ntpdate
7 Sep 21:46:04 ntpdate[3909]: no servers can be used, exiting
[root@frontend ~]# ntpdate 213.203.238.82
7 Sep 21:46:11 ntpdate[3910]: step time server 213.203.238.82 offset -0.872507 sec
[root@frontend ~]#
Maybe this is also important.
Thanks!