[ntp:questions] drift value very large and very unstable

Andy Helten andy.helten at dot21rts.com
Fri Mar 7 19:47:52 UTC 2008


Miroslav Lichvar wrote:
> On Fri, Mar 07, 2008 at 09:13:14AM -0600, Andy Helten wrote:
>   
>> As I mentioned in a previous email, I was going to run ntp while adding
>> back in some of the features I removed.  After doing this, at least
>> superficially, I've isolated the problem to time step being disabled. 
>> It doesn't matter whether I specify 'tinker step 0' in ntp.conf or use
>> the '-x' argument on the command line.
>>     
>
> This looks like a kernel bug. When the step threshold is 0 or larger
> than 0.5s, kernel time discipline is disabled and time is adjusted
> only by adjtime(). I've recently seen a similar problem on a PowerPC
> machine where adjtime() called with delta smaller than 1ms was
> ignored.
>   

Thank you for this pointer!  I'm fuzzy at best on the relationship
between NTP and the kernel, but I did find a read through a thread on
the mailing list about 'tinker step 0' and the kernel time discipline:

http://lists.ntp.isc.org/pipermail/questions/2006-September/011531.html

I read through it quickly, so I am still unsure why disabling the kernel
time discipline is necessary if stepping is disabled but no step would
have occurred even if stepping were enabled.  Is there an "official" NTP
page that covers the topic of kernel time discipline?  At any rate, I
will look closer at the kernel and the problems with adjtime.

Andy





More information about the questions mailing list