[ntp:questions] Can a clock drift be too big for ntpd?

Steve Kostecke kostecke at ntp.org
Sat Oct 20 02:34:31 UTC 2007


On 2007-10-19, Patrick Nolan <pln at glast2.Stanford.EDU> wrote:

> On 2007-10-19, Hal Murray wrote:
>
>> Do you have any other similar machines?
>
> We bought a batch of 4 identical Dell Optiplex 745 machines. Three are
> running Linux and one Windows. Only one has this problem.

It is possible that you have one motherboard with a hardware problem.

>> Are you missing interrupts? (say because some strange IO device eats
>> up a lot of CPU processing its interrupts)
>
> I'm not sure how to look for this.

It is de-rigueur to mention Linux and missing interrupts in the same
sentence. Even in the absence of evidence.

If you can establish a correlation between clock resets and periods of
heavy IDE disk activity, you may have an interrupt problem.

> I realized something since yesterday. This machine's unique feature
> is that I compiled a new kernel to add the Reiser file system. I
> downloaded the source RPM for the exact version that came with the
> distro, and the only thing I changed was to enable ReiserFS.

Can you try recompiling the kernel without Reiser FS and seeing if that
makes a difference?

On a 2.6.15 kernel the HZ setting may be found in "make menuconfig"
under "Processor type and features" as "Timer frequency".

FWIW, my 2.6.15 kernel is set to 250Hz and ntpd has no problem with it.

-- 
Steve Kostecke <kostecke at ntp.org>
NTP Public Services Project - http://support.ntp.org/




More information about the questions mailing list