[ntp:questions] no drift-file on 2008 R2 vps and the time diff. is getting bigger and bigger?

William Unruh unruh at invalid.ca
Sun Sep 14 09:00:49 UTC 2014


On 2014-09-14, gooly <gooly at gmx.at> wrote:
> Thank you, Harlan
> now I start to understand!
>
> (keeping all is getting too long, so I delete most of the post)
>
>> The frequency of 500ppm indicates ntpd has "hit its limit" of how much
>> adjusting it can do.  leap=11 means ntpd is not synchronized.
>>
>> The first thing to fix is the problem with clock frequency.
> I can't fix the clock freq.-problem. It's a vps which probably shares 
> the physical server with other vps.

And furthermore has no requirement that the time it provides has any
relation to real time. The software should be set up so that the time in
the 'VPS' is derived from the clock running onthe background system. It
is that background server that should be running ntp. The VPS itself
will have extreme difficulty in keeping good time itself since most of
the clock it is not running at all. 
> Is this a valid indicator that the provider should do a better load 
> distribution because other run on their vps programs at 'full power' 
> which 'steals' the clock interrupts?
> I guess I have to switch the provider.

It is an indicator that the time within virtual server should be
delivered by the background hardware, not from within the VPS itself. 

>
> 2)
>> If you run at debug level 1, you should also see messages about the
>> drift file update process.
> How do I start the Windows-Meinberg-version in a debug modus?
> The cmd-start-line is:    net start ntp

And then as well you are running windows which, even when it directly
controls the hardware has trouble giving good time. 

>
> 3)
>> On some systems (as I recall), a non-existent drift file will not be
>> created - you might try creating an empty one and seeing what happens.
> I took the one of my pc and copied it into the etc-folder and changed 
> its value to 500 - I'll see.
>
> 4)
>> If yuor system is *slowing* by 7 sec every 15 minutes, I suspect you are
>> losing clock interrupts.
>>
>> At that rate ntpd will never be able to keep your clock sync'd.
> Does this mean ntp will not even try to set the clock - it just gives up?

Yes. If the drift is greater than 500 it gives up. 

>
> I just restarted ntp by "net start ntp -g" the clock was adjusted and 
> now has a offset of 0.33 sec,
> the ntp.drift exitst in /etc with '500'.
> I'll see what happens.

You are like someone who is trying to build a high accuracy astronomical
observatory on top of a swamp. 



More information about the questions mailing list