ntp service causing huge time errors

zoubidoo · 11-23-2004, 06:52 AM

I recently started using the ntpd service to keep my clock accurate.
When I start up the service, the time synchronises fine, but afterwards I get huge errors (several hours in less than a day). If I restart the ntpd service, the time is once again accurate. What is causing the time to go so badly wrong?

Any suggestions much appreciated,

PJ

Platform: FC2, 2.6.8-1.521, ntp-4.2.0-7
Here's my log output:

Code:

[root@darwin zbd]# tail /var/log/ntp.log
23 Nov 12:02:24 ntpd[1025]: running as uid(38)/gid(38) euid(38)/egid(38).
23 Nov 12:36:40 ntpd[1025]: synchronized to 195.13.1.153, stratum=2
23 Nov 12:50:05 ntpd[1025]: time reset +805.706758 s
23 Nov 12:50:05 ntpd[1025]: kernel time sync disabled 0041

Here's the /etc/ntp.conf:

Code:

restrict 127.0.0.1
fudge 127.127.1.0 stratum 3
server europe.pool.ntp.org
driftfile /var/lib/ntp/drift
broadcastdelay  0.008
logfile /var/log/ntp.log

Here's /var/log/ntp/drift:

Code:

500.000

Technoslave · 11-23-2004, 10:11 AM

Weird. A long time ago, when I first started using linux the ntpd that was included didn't work very well, so I ended up making my own script to update the clock. These problems have since been fixed and on all the boxes I admin ntpd or xntpd works fine. Here is my script that I run on my home linux box, to this day:

#!/bin/sh

CLOCKSERVER="clock.psu.edu"
CUSER=`echo $UID`

if [ $CUSER -ne 0 ]
then
echo "You must be root in order to execute this script"
exit 1
fi

/usr/sbin/ntpdate $CLOCKSERVER

# Setting hardware clock to new time

/sbin/hwclock --systohc

***

I have this run every hour out of root's crontab.

***

Anyway, just a solution, not an answer to your problem.

zoubidoo · 11-23-2004, 10:29 AM

Quote:

Originally posted by Technoslave
[snip]
Anyway, just a solution, not an answer to your problem.

Thanks, if all else fails I'll borrow this script. Of course, it'd be nice to fix the problem though...

PJ