[mythtv-users] Zotac ION + NTP clock too fast?

[Matt W]

On Fri, Oct 16, 2009 at 1:14 PM, Brian Long <briandlong at gmail.com> wrote:
> I've noticed my Zotac IONITX-C frontend is configured to use 3 NTP servers,
> but it's time is increasing too quickly.  Last night, I disabled NTP, ran
> ntpdate a few times, verified it was the same time as my MBE, then started
> NTP again.  By this morning, it was 3 minutes ahead.  Has anyone else
> noticed their ION box running too fast?
>
> The ION is PXE-booting Fedora 11 i386 and NFS mounting root over NFS.  The
> /var/lib/ntp/drift file is part of the host-specific snapshot directory and
> is writable.  I've not yet enabled ntpd debug messages to see if I can
> narrow down the problem.  Thanks for any feedback.

I would go through your ntp.conf file and rip out a lot of the
'cruft'.  I still don't understand why so many Linux distros' feel the
need to put the LOCAL clock as a peer in the stock config.  Rip that
out if it's there.

re: logging, you can set up a basic log structure that will tell you a
lot without flooding your disk with junk messages.  I typically use:

logfile /var/log/ntp/ntpd.log
logconfig +sysall
logconfig +peerinfo -peerevents +peerstatus
logconfig +syncall
logconfig +clockall

obviously clockall is not needed if you're not running a reference
clock so you can probably ditch that line.

to "clean your clock", first remove the drift file completely.  Then
disable any autostart of your ntp stuff.  Reboot the box, and run
'ntpdate -bsv <somentphost>'.  The -b flag will force a hard step
correction instead of a slew.  Then start your normal ntpd startup
script.  or modify your startup script to use a -b flag on the ntpdate
portion.

any chance your mobo coin cell / lithium battery is old/flaky ?

good luck and chime on!  :-)