Hello,

I have a Linux server running CentOS 4.3 32 bit with the 2.6.9-34.0.2.ELsmp kernel. It has a Supermicro PDSMi motherboard with a Celeron D 3.06GHz processor. The board uses APIC for interrupts as it can handle Pentium D (dual-core) chips so I am running the SMP kernel to handle that even though the Celeron is only 1 logical processor.

The system clock is always running fast, gaining several minutes a day that it shouldn't and I'm trying to figure out why, and how to fix it. If I run the server without NTP, the system clock will continually run fast but the hardware clock will run normal (hwclock --show).

If I run NTP (pointing at pool.ntp.org servers), both the system and hardware clock run several minutes faster than they should. For example, see below:

[root@neon ~]# date ; hwclock --show
Fri Jul 28 20:16:52 EDT 2006
Fri 28 Jul 2006 08:16:52 PM EDT -0.099188 seconds
[root@neon ~]# ntpq -p
remote refid st t when poll reach delay offset jitter
==============================================================================
goldengate.vpiz 204.123.2.5 2 u 576 1024 377 84.742 -362748 8053.65
mail.confluxind 130.126.24.44 3 u 577 1024 377 39.623 -362746 8074.40
broken.blackros 24.130.58.99 2 u 646 1024 377 1.841 -362188 8049.68
*LOCAL(0) LOCAL(0) 10 l 58 64 377 0.000 0.000 0.001
[root@neon ~]# /etc/init.d/ntpd stop
Shutting down ntpd: [ OK ]
[root@neon ~]# ntpdate 0.pool.ntp.org
28 Jul 20:11:43 ntpdate[2599]: step time server 69.3.12.50 offset -367.734784 sec
[root@neon ~]# date ; hwclock --show
Fri Jul 28 20:11:50 EDT 2006
Fri 28 Jul 2006 08:17:58 PM EDT -0.930095 seconds
[root@neon ~]# /etc/init.d/ntpd start
Starting ntpd: [ OK ]
[root@neon ~]#


So, the clock is five minutes faster than it should be. If I stop NTP and manually sync, it'll go back to the correct time. However, by the end of the day, it'll be off and continue to drift. One thing I noticed is the 8000+ jitter. That seems extremely excessive. Is that what is causing this? Also, why would the system clock run so fast even when NTP is not running. I'm not much of an expert on NTP, any help with this would be appreciated. Am really trying to determine if it is a hardware problem with the HW clock, though that seems okay, or something that can be fixed via the software.

Thanks for any help, I'll let all know if I find anything.
Tom

Looks like your ntpd is not connecting. You need a single line in ntpd.conf: servers pool.ntp.org

Thanks for the reply. It turns out that NTP was connecting but then going into a rejected status for each of the various time sources due to the drift being to great to accomodate a sync. It appears that the RTC chip on the motherboard is bad, causing the system clock to speed excessively. I have received a new board via RMA and will set that up in the future to see if the behavior continues.

I also configured a system with the same OS, kernel and motherboard and am having no issues with NTP, so am fairly certain that this is/was a hardware related issue.

Best Regards,
Tom