NTP problem - Clock too fast for NTP to keep up?

[Matthew Saltzman]

On Thu, 10 Feb 2005, Peter Kiem wrote:

> Hi Gene,
>
>> Given a stable clock in your box, which from the reports below I doubt, 
>> every 2 minutes would, I think, give results that are totally lost in the 
>> propagation delays of the net.
>
> Actually the physical server is fine, only need to run ntpdate once per hour 
> on it.
>
> And this is contacting my firewall for the time, not the net.  I have other 
> servers synchronising off the same firewall without issues.
>
> Really seems like something screws up running Fedora inside VMware GSX.
> Everything else is running perfectly though, just the system clock is screwy. 
> Windows in a virtual server on the same host server is fine.

I haven't been following this thread closely, but the ref to VMware caught 
my eye.  In VMware WS, there is an option in VMware Tools to synch the 
VM system clock to the host system clock.  (I've only run Windows 
guests, so I'm not sure where to look for a Linux guest.)  I'd be 
surprised if GSX doesn't have a similar option.  Do you have it set?

>
>> This doesn't look like a tickadj correctable error to be truthfull unless 
>> you take tickadj down in the 6000 - 7500 range.  This one 
>
> Well I'm slowly reducing it (currently at 9900) and this is what I am seeing 
> now.
>
> Feb 10 16:47:58 krusty ntpdate[9134]: step time server 202.173.151.129 offset 
> -3.756325 sec
> Feb 10 16:49:52 krusty ntpdate[9146]: step time server 202.173.151.129 offset 
> -9.375502 sec
> Feb 10 16:50:01 krusty ntpdate[9153]: adjust time server 202.173.151.129 
> offset -0.086756 sec
> Feb 10 16:51:51 krusty ntpdate[9166]: step time server 202.173.151.129 offset 
> -9.705495 sec
> Feb 10 16:52:02 krusty ntpdate[9172]: adjust time server 202.173.151.129 
> offset -0.456918 sec
>
> Looks quite promising, it's a lot closer than before.
>
>> Asking the next obvious question, do you have cpufreq enabled?
>
> Um nope.  No program found by that name although I did find cpufreq in the 
> kernel modules but it isn't a loaded module according to lsmod.
>
> There is also a cpuspeed ticked on in ntsysv so could that be an issue?
>
>

-- 
 		Matthew Saltzman

Clemson University Math Sciences
mjs AT clemson DOT edu
http://www.math.clemson.edu/~mjs