[CentOS] Strange NTP problem

[Filipe Brandenburger]

On Tue, May 20, 2008 at 8:53 PM, Alfred von Campe <alfred at von-campe.com> wrote:
> # ntpq -np
>     remote           refid      st t when poll reach   delay   offset
>  jitter
> ==============================================================================
>  10.101.32.104   67.128.71.65     3 u  689 1024  377    0.659  54095.7
> 4263.68
> *127.127.1.0     .LOCL.          10 l   28   64  377    0.000    0.000
> 0.001
>
> The bad system is synch'ing with itself (which was apparent from the
> messages in /var/log/messages).  The question is why?

When the offset is too big, NTP will discard the server as a source of
synchronization. Try stoping ntpd, setting the date with ntpdate, then
starting ntpq -np again to see what happens.

I saw that the jitter appears to be really big too. That might be
caused by your server going too slow or too fast in a rate that NTP
couldn't keep up to. Once it happened to me, somehow (IRQ problems?)
one machine started going so fast that NTP couldn't keep up with it,
so I had to put periodic ntpdate's to keep it at the right time. I
rebooted it off-hours and the problem went away. Did you try rebooting
this machine and seeing if it fixes the problem?

One good side is the "reach = 377", that means your network is good
and all NTP packets are being exchanged correctly. 377 is octal for
binary 11111111, which means the last 8 NTP packet exchanges were
successful.

Good luck!
Filipe