f



losing time

I'm running rh8 kernel 2.4.20 on a Dell Inspiron 8000 P3 800 laptop w/ 
ntp version 4.1.1

I have been consistently losing about 19.5 seconds every 16.5 minutes, 
at which time ntp resets the time with the time of my local server.  It 
seems to me that with a consistent error such as this it should be able 
to correct itself, but it hasn't been able to.

As per Harlan's suggestion I have tried running with the '-N high' 
option, but i'm still getting a consistent error.

Any suggestions?

    Brendon
    brendon@iqon.net

0
Brendon
7/17/2003 10:55:56 PM
comp.protocols.time.ntp 4895 articles. 2 followers. Post Follow

5 Replies
568 Views

Similar Articles

[PageSpeed] 41

I turned off ntpd and I lost 133 seconds in 50 minutes (loss of .375 
seconds per minute) which was better than ntpd's .846 seconds per minute.

    Brendon

Frederick Bruckman wrote:
> In article <bf799n$ieo$1@sun-news.laserlink.net>,
> 	Brendon Mosher <brendon@iqon.net> writes:
> 
>>I'm running rh8 kernel 2.4.20 on a Dell Inspiron 8000 P3 800 laptop w/ 
>>ntp version 4.1.1
>>
>>I have been consistently losing about 19.5 seconds every 16.5 minutes, 
>>at which time ntp resets the time with the time of my local server.  It 
>>seems to me that with a consistent error such as this it should be able 
>>to correct itself, but it hasn't been able to.
>>
>>As per Harlan's suggestion I have tried running with the '-N high' 
>>option, but i'm still getting a consistent error.
> 
> 
> That's a pretty severe error. What happens when you don't run "ntpd"? 
> Maybe the problem's in the kernel, and not "ntpd" at all.
> 

0
Brendon
7/18/2003 6:56:07 AM
In article <bf85ib$jp6$1@sun-news.laserlink.net>, brendon@iqon.net says...
>
>I turned off ntpd and I lost 133 seconds in 50 minutes (loss of .375 
>seconds per minute) which was better than ntpd's .846 seconds per minute.
>
>    Brendon
>
>Frederick Bruckman wrote:
>> In article <bf799n$ieo$1@sun-news.laserlink.net>,
>>       Brendon Mosher <brendon@iqon.net> writes:
>> 
>>>I'm running rh8 kernel 2.4.20 on a Dell Inspiron 8000 P3 800 laptop w/ 
>>>ntp version 4.1.1
>>>
Laptops aren't meant for time keeping. Somehow the energy saving functions 
of the laptop interfere with the clock. But anyway, I use a laptop as 
primary ntp-server. At first I got similar errors as you do. Later I happened
to notice that running xosview helped (currently the offset to both upstream
NTP-servers is 5 msec). 

[In my experiments, xosview is the only program that helps. I tried al kinds of 
mock-programs, just reading timers in /proc eg. Nothing helped.
I currently use Xvnc as X-server, so I don't need a real display.]

Succes, Johan

0
J
7/18/2003 7:35:32 AM

Brendon Mosher wrote:
> 
> I turned off ntpd and I lost 133 seconds in 50 minutes (loss of .375
> seconds per minute) which was better than ntpd's .846 seconds per minute.

Run those numbers by me again?

133 secs in 50 mins -> 2.66 secs per min (but .375 = 1 / 2.66)
19.5 secs in 16.5 mins -> 1.18 secs per min (and .846 = 1 / 1.18)

So with NTP it is better than without, not worse, but still pretty
awful.

Either way, a basic clock error of 44333 ppm is a ludicrously bad clock!
The NTP is probably pulling the clock to the limit the OS allows
(apparently about 1.5 secs per min), but nothing can compensate for that
kind of gross clock error.

John Howells

> 
>     Brendon
> 
> Frederick Bruckman wrote:
> > In article <bf799n$ieo$1@sun-news.laserlink.net>,
> >       Brendon Mosher <brendon@iqon.net> writes:
> >
> >>I'm running rh8 kernel 2.4.20 on a Dell Inspiron 8000 P3 800 laptop w/
> >>ntp version 4.1.1
> >>
> >>I have been consistently losing about 19.5 seconds every 16.5 minutes,
> >>at which time ntp resets the time with the time of my local server.  It
> >>seems to me that with a consistent error such as this it should be able
> >>to correct itself, but it hasn't been able to.
> >>
> >>As per Harlan's suggestion I have tried running with the '-N high'
> >>option, but i'm still getting a consistent error.
> >
> >
> > That's a pretty severe error. What happens when you don't run "ntpd"?
> > Maybe the problem's in the kernel, and not "ntpd" at all.
> >
0
John
7/18/2003 7:45:18 AM
This is a machine I dual boot with windows.  Under windows it never 
loses any time, is this something that windows is correcting?  If they 
can correct it there shouldn't I be able to correct under linux?

(I guess my math was a little off there :), it was getting kind of late)

    Brendon

Johan Swenker wrote:
> In article <bf85ib$jp6$1@sun-news.laserlink.net>, brendon@iqon.net says...
> 
>>I turned off ntpd and I lost 133 seconds in 50 minutes (loss of .375 
>>seconds per minute) which was better than ntpd's .846 seconds per minute.
>>
>>   Brendon
>>
>>Frederick Bruckman wrote:
>>
>>>In article <bf799n$ieo$1@sun-news.laserlink.net>,
>>>      Brendon Mosher <brendon@iqon.net> writes:
>>>
>>>
>>>>I'm running rh8 kernel 2.4.20 on a Dell Inspiron 8000 P3 800 laptop w/ 
>>>>ntp version 4.1.1
>>>>
> 
> Laptops aren't meant for time keeping. Somehow the energy saving functions 
> of the laptop interfere with the clock. But anyway, I use a laptop as 
> primary ntp-server. At first I got similar errors as you do. Later I happened
> to notice that running xosview helped (currently the offset to both upstream
> NTP-servers is 5 msec). 
> 
> [In my experiments, xosview is the only program that helps. I tried al kinds of 
> mock-programs, just reading timers in /proc eg. Nothing helped.
> I currently use Xvnc as X-server, so I don't need a real display.]
> 
> Succes, Johan
> 

0
Brendon
7/18/2003 5:39:49 PM
For those interested, I did track down the problem.

It is laptop related problem that has apparently plagued many Dell 
laptops, and a fix was only recently made available.  I was running the 
gnome battstat applet, each request it made to monitor the battery 
slowed down the clock.  A firmware upgrade made available by Dell for
the Inspiron 8000 solved the problem.

Thanks for the Help!
    Brendon
    brendon@iqon.net

Brendon Mosher wrote:
> I turned off ntpd and I lost 133 seconds in 50 minutes (loss of .375 
> seconds per minute) which was better than ntpd's .846 seconds per minute.
> 
>    Brendon
> 
> Frederick Bruckman wrote:
> 
>> In article <bf799n$ieo$1@sun-news.laserlink.net>,
>>     Brendon Mosher <brendon@iqon.net> writes:
>>
>>> I'm running rh8 kernel 2.4.20 on a Dell Inspiron 8000 P3 800 laptop 
>>> w/ ntp version 4.1.1
>>>
>>> I have been consistently losing about 19.5 seconds every 16.5 
>>> minutes, at which time ntp resets the time with the time of my local 
>>> server.  It seems to me that with a consistent error such as this it 
>>> should be able to correct itself, but it hasn't been able to.
>>>
>>> As per Harlan's suggestion I have tried running with the '-N high' 
>>> option, but i'm still getting a consistent error.
>>
>>
>>
>> That's a pretty severe error. What happens when you don't run "ntpd"? 
>> Maybe the problem's in the kernel, and not "ntpd" at all.
>>
> 

0
Brendon
7/22/2003 3:19:38 PM
Reply: