On Fri, 16 Jun 2006, in the Usenet newsgroup comp.protocols.time.ntp, in article
<[EMAIL PROTECTED]>, Jeremy Parrish wrote:
>We just bought 7 new servers from SuperMicro, and they all have the same
>problem. They all gain about 1 second every 10 seconds or so (10% too
>fast). I have tried various versions of Linux kernels (in particular,
>2.4.27, 2.6.15, 2.6.15-smp, 2.6.16.20...) with all kinds of kernel
>options (noapic, acpi=off, clock=tsc, the list goes on...). There is no
>noticeable difference between any of those.
You have something horribly mis-configured in the BIOS setup - I have
no idea what it might be. See what your timer interrupts are looking like.
Get a real watch, and enter the following command as close to 100 seconds
apart as you can manage
[compton ~]$ grep ' 0:' /proc/interrupts
0: 1022180577 timer
[compton ~]$
Depending how your kernel is compiled, you should see 100 or 1000 interrupts
per second - not 111 or 1111. Also look at the boot messages and see what
it's claiming for BogoMips.
>Is there a way I can make ntpd play nice with such a horrible drift?
A ten percent error is not a bad crystal (certainly not on seven systems),
when the average crystal in PCs is 50 or 100 ppm over temperature and
voltage ranges. Get the BIOS problem fixed before thinking about time
protocols of ANY kind.
Old guy
_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.isc.org/mailman/listinfo/questions
