Time running too fast - uhm??
Alexander Holst
yellowdog-general@lists.terrasoftsolutions.com
Thu Jan 30 07:52:01 2003
Hi List,
I did post this issue a few days ago, but never saw any reaction to it.
I still find my log file on only _one_ particular machine full of
entries like:
Jan 30 09:36:08 wwwsvr02 ntpd[5356]: time reset -1.123347 s
Jan 30 09:36:08 wwwsvr02 ntpd[5356]: synchronisation lost
Jan 30 09:40:33 wwwsvr02 ntpd[5356]: time reset -0.856180 s
Jan 30 09:40:33 wwwsvr02 ntpd[5356]: synchronisation lost
Jan 30 09:49:16 wwwsvr02 ntpd[5356]: time reset -1.427873 s
Jan 30 09:49:16 wwwsvr02 ntpd[5356]: synchronisation lost
Jan 30 09:54:49 wwwsvr02 ntpd[5356]: time reset -0.965880 s
Jan 30 09:54:49 wwwsvr02 ntpd[5356]: synchronisation lost
Jan 30 10:03:39 wwwsvr02 ntpd[5356]: time reset -1.282842 s
Jan 30 10:03:39 wwwsvr02 ntpd[5356]: synchronisation lost
Jan 30 10:14:23 wwwsvr02 ntpd[5356]: time reset -0.460586 s
Jan 30 10:14:23 wwwsvr02 ntpd[5356]: synchronisation lost
Jan 30 10:18:50 wwwsvr02 ntpd[5356]: time reset -1.657446 s
Jan 30 10:18:50 wwwsvr02 ntpd[5356]: synchronisation lost
My /etc/ntp/drift looks like:
[user@host user]$ cat /etc/ntp/drift
-262.668
I have never seen values like that in a drift file. My question is: why
is the systemtime running that fast?
Funny enough, the hardware clock runs pretty normal as you can see:
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:22 PM CET -0.276448 seconds
Thu Jan 30 15:46:21 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:36 PM CET -0.610453 seconds
Thu Jan 30 15:46:35 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:39 PM CET -1.010177 seconds
Thu Jan 30 15:46:38 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:42 PM CET -0.970197 seconds
Thu Jan 30 15:46:41 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:44 PM CET -0.297311 seconds
Thu Jan 30 15:46:43 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:48 PM CET -0.072218 seconds
Thu Jan 30 15:46:47 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:50 PM CET -0.361500 seconds
Thu Jan 30 15:46:49 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:52 PM CET -0.623877 seconds
Thu Jan 30 15:46:51 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:54 PM CET -0.151135 seconds
Thu Jan 30 15:46:53 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:57 PM CET -0.848402 seconds
Thu Jan 30 15:46:56 CET 2003
[holale@wwwsvr02 holale]$ hwclock -r; date
Thu 30 Jan 2003 04:46:59 PM CET -0.802279 seconds
Thu Jan 30 15:46:58 CET 2003
The gap increases quite noticably, untill ntpd tries to slew the time
again - but somehow can't keep up with that, so roughly every five
minutes it has to force the time back.
I have several other Macs (G3, 7500, ANS) that do not show this
behaviour (the machine in question is a PM8200). What parameters are
responsible for keeping the systemtime accurate? What makes the machine
miscalculate the systemtime, so ntpd has to reset the clock almost
every five minutes?
If it wasn't a host that receives logging info from other hosts, I
wouldn't mind the time issue.
Any ideas appreciated,
Alex
Alexander Holst
Pforzheim University of Applied Sciences
<holst@fh-pforzheim.de>
ph: +49 [0]7231 28-6837
fx: +49 [0]7231 28-6040