[CentOS] ntpd

Wed Dec 12 14:51:21 UTC 2007
Brian Mathis <brian.mathis at gmail.com>

On Dec 12, 2007 12:50 AM, Jason Pyeron <jpyeron at pdinc.us> wrote:
> I am running a server inside of VMWare, and the clock gains ~30 seconds
> every 1000 seconds or 1.03X.
>
> I need to keep the drift under the magic 1000 limit that ntpd kills its
> self, but despite setting maxpoll really low I get:
>
> Dec 11 23:58:14 host ntpd[4909]: kernel time discipline status change 41
> Dec 11 23:59:17 host ntpd[4909]: kernel time discipline status change 1
> Dec 11 23:59:17 host ntpd[4909]: time correction of -1123 seconds exceeds
> sanity limit (1000); set clock manually to the correct UTC time.
>
>
> /etc/ntp.conf:
>
> server time.intranet.pdinc.us maxpoll 7
>
> Ideas? If I cannot get ntpd working, then I will have to resort to a cron *
> * * * * rdate -s time.intranet.pdinc.us
>


Do not use ntp to sync time on the guest OS.  Sync using ntp on the
HOST, and then use the vmware tools to sync in the guest.  I have
pursued this issue many times, and that is the best answer.

Then update your kernel boot parameters and add:
    clock=pit              # for kernels less than 2.6
OR
    clocksource=pit    # for kernels 2.6.16 and later

There is far more than you ever wanted to know about vmware time syncing here:
    http://www.vmware.com/pdf/vmware_timekeeping.pdf