[ntp:questions] NTP 4.2.6p4 Released

unruh unruh at wormhole.physics.ubc.ca
Thu Sep 29 05:40:01 UTC 2011

On 2011-09-28, David J Taylor <david-taylor at blueyonder.co.uk.invalid> wrote:
> "unruh" <unruh at wormhole.physics.ubc.ca> wrote in message 
> news:slrnj86n3g.emt.unruh at wormhole.physics.ubc.ca...
> []
>> Is he using a virtual system? Is that loss constant ( or is it sometimes
>> 1 sec, sometimes 3 sec sometimes nothing). He is eitehr running a
>> virtual machine, in which case he should not be using ntpd, or his
>> operating system is doing an attrocious job of setting the system time
>> rate of his clock.
> Thanks for your thoughts, Bill.  No, this is a real PC, which used to work 
> correctly.  There is a question as to (a) whether the PC has become 
> faulty, (b) whether something had changed in the software, or (c) whether 
> some other program is interfering with the time.  Nothing is supposed to 
> have changed in the software, and there are no restore points early enough 
> to check (b).  The obvious candidates for (c) are not present, which 
> leaves (a) as the obvious cause, but I has my doubts that a PC would 
> suddenly start to lose 1 second every ten minutes.
> I don't know whether the loss is constant, as with NTP running it keeps 

No way. ntpd will not reset the clock. If it finds the rate that far
out, it will shut down. 

> resetting, and investigations are continuing.  It's at the point where I'm 
> running out of ideas, so I may, very reluctantly, ask Dave H to get 
> involved, if the current tests prove unsuccessful.  These tests are now to 
> try and run as plain an NTP as possible, but a more recent version 
> (4.2.6p4).  I'll post the outcome here.

The question is whether or not it is consistant, in which case it sounds
like the system is not setting the intial clock rate properly. The
system has an internal crystal, and a timer chip which counts a certain
number of oscillations and interrupts. It also calibrates the cpu
counter against the timer ticks. Now it is hard to imagine how that
hardware would go wrong. 

a)find out if the time rate loss is consistant. 
b) Try running chrony instead of ntpd and find out what the rate
correction is and see if it is consistant. 

> Cheers,
> David 

More information about the questions mailing list