[BlueOnyx:04083] Re: NTP etc

[Chris Gebhardt - VIRTBIZ Internet]

Hi Jason,

Jason Ozin wrote:
> And to further answer and refine my answer this is what was needed in
> crontab
> In case anyone is interested, this clock drift is a known issue when
> virtualising BlueOnyx/Linux on VMWare or Hyper-V 
> 
> # update clock every 30 mins
> 10,40 * * * * /usr/sbin/ntpdate -s -b -u pool.ntp.org
> # sync hardware clock 5 mins later
> 15,45 * * * * /sbin/hwclock --systohc

Hmmm, actually, the drift is not really due to VMware or Hyper-V, but 
rather the way that the guest OS responds to the virtualized clock.  And 
running the ntpdate every 30 minutes is, at best, a big hammer used to 
make a fine adjustment.  Realistically, it's not a fix at all.

If you would rather fix the problem, then first determine if the clock 
runs fast or slow.   My gut is that your guest OS has a fast clock, 
since that is what I have most often encountered with most modern 
multi-core systems (especially AMD, but also in Intel when the 
power-management features are BIOS enabled).

The fix here is to pass the following kernel arguments:
processor.max_cstate=1 clock=pmtmr divider=10

Placing those arguments into the kernel line of your grub boot file 
(/boot/grub/menu.lst) should compensate.

Remember to add it to the top area of of the file as well (which is 
commented with #) so that when a new kernel is installed it will carry 
those arguments.

That will help your guest OS actually keep time, rather than having to 
be beat over the head every 30 minutes and still drift.
-- 
Chris Gebhardt
VIRTBIZ Internet Services
Access, Web Hosting, Colocation, Dedicated
www.virtbiz.com | toll-free (866) 4 VIRTBIZ