Topic: Problems with time drift, hardware clock, and ntp - solved?

[holck]

Last week I noticed that the time on my small server was off by several hours, even though I have ntp configured and running. I corrected the time, restarted ntp, and hoped that everything was ok. But the next day, the clock was wrong again. I /var/log/messages I found a few line like this

Code: [Select]

May 30 21:59:59 katrine kernel: set_rtc_mmss: can't update from 0 to 59
I I am certainly not an expert in ntp, so I spent a lot of time trying to figure out what was going on, and how to solve it. I got a lot of help from these:

• http://www.math.ucla.edu/~jimc/documents/bugfix/12-ntp-wont-sync.html
• http://log.or.cz/?p=80

Using ntpq I could see that my clock was drifting crazily - several seconds per minute. As I understood it, I had some problems with my hardware clock. I installed the package adjtimex, and by removing /etc/adjtime and /etc/ntp drift, and using the programs tickadj and adjtimex, I finally managed to get ntpd up, running and synchronizing with remote ntp servers again.

Right now, ntpq -p shows

Code: [Select]

     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
+loin.ploup.net  88.191.254.6     3 u  822 1024  377   24.987    0.473   0.721
+server01.coloce 46.19.33.5       3 u  872 1024  377   19.108   -1.275   0.034
-37.247.49.133   64.90.182.55     2 u  672 1024  377   26.183   10.493   0.250
*ev003.tilaa.nl  193.67.79.202    2 u  143 1024  377   16.690   -0.456   0.029
so this looks fine, I think. But /var/log/ntpd/current shows

Code: [Select]

2013-06-01 22:23:23.577990500  1 Jun 22:23:23 ntpd[12861]: Listening on interface wildcard, 0.0.0.0#123 Disabled
2013-06-01 22:23:23.578033500  1 Jun 22:23:23 ntpd[12861]: Listening on interface lo, 127.0.0.1#123 Enabled
2013-06-01 22:23:23.578074500  1 Jun 22:23:23 ntpd[12861]: Listening on interface eth1, 5.103.132.90#123 Enabled
2013-06-01 22:23:23.578113500  1 Jun 22:23:23 ntpd[12861]: Listening on interface eth0, 192.168.10.1#123 Enabled
2013-06-01 22:23:23.578171500  1 Jun 22:23:23 ntpd[12861]: kernel time sync status 0040
2013-06-01 22:23:24.152584500  1 Jun 22:23:24 ntpd[12861]: frequency initialized 0.000 PPM from /etc/ntp/drift
2013-06-01 22:26:39.585026500  1 Jun 22:26:39 ntpd[12861]: synchronized to 46.19.33.6, stratum 2
2013-06-01 22:26:39.585078500  1 Jun 22:26:39 ntpd[12861]: kernel time sync disabled 0001
2013-06-03 01:02:31.597177500  3 Jun 01:02:31 ntpd[12861]: synchronized to 37.247.49.133, stratum 2
2013-06-03 01:21:49.595153500  3 Jun 01:21:49 ntpd[12861]: synchronized to 46.19.33.6, stratum 2

and I'm wondering about the "kernel time sync disabled" message. Should I worry about this? Should I just hope that everything is fine now, or do I need to do something about my hardware clock?

Thanks, Jesper

[Stefano]

is it a virtual machine o a physical one?

[holck]

It is physical, a small HP Proliant server

[Stefano]

never had such an issue.. and I have several hp server working with SME..

what about HP customer care? seems like a machine bug to me..

[larieu]

I had something like this before
SME 7.x working ok
after almost 2 years of working ok one update of kernel for SME arrived
after that the clock get crazy

http://forums.contribs.org/index.php/topic,42831.msg202622.html#msg202622

the problem was the cipset (INTEL MB original) which wasn't supported any more by CentOS
verify the cipset vs the known clock related problems on CentOS