[ntp:questions] Hopelessly broken clock?

Unruh unruh-spam at physics.ubc.ca
Mon Nov 24 03:27:43 UTC 2008

Chris Richmond <tomnykds at comcast.net> writes:

>Harlan Stenn <stenn at ntp.org> wrote in
>news:ywn9y6zgwc38.fsf at ntp1.isc.org: 

>>>>> In article <Xns9B5ABA19BFBAtomnykdscomcastnet at>, Chris
>>>>> Richmond <tomnykds at comcast.net> writes: 
>> Chris> /usr/sbin/ntpq -p -n
>> Chris> remote       refid   st t when poll reach delay offset jitter
>> Chris>
>> =======================================================================
>> = Chris> * .PPS.    1 u 94    128   377 0.193 -0.837 5.872
>> Chris> .LOCL.  10 l 16     64   377 0.000 0.000  0.000
>> So why are you running a local refclock on this box?

>Only because it was in the ntp.conf as a "just in case no other sources
>are available" comments.  There's no plan to leave it in.  Gone now.

Yes, many distros do that. It makes no sense. If no other sources are
available, gazing at your own navel is not going to get you better time. 
ONLY if your system is a server for other machines, and it is critical that
those other machines always think that your machine is synchronized with
something ( eg they have no other servers than your machine) does it make
sense to have the local refclock.

>Still can't get around SELinux blocking the writing of the drift file.
>Nov 19 12:52:37 router1 ntpd[2348]: can't open /etc/ntp/drift.TEMP: 
>Permission denied

create it and make it read-writable by all.

>So, I moved the drift file to /root.  So far, so good, but no drift file
>after ~15 minutes, but no log messages either.  Nowhere near the accuracy
I think it is once an hour the drift file is written.

>of the reference machine, but at least it's working.

>[crichmon at router1 ~]$ /usr/sbin/ntpq -p -n
>remote           refid      st t when poll reach   delay   offset  jitter
>*     .PPS.       1 u   17   64  377    0.206  -23.479   4.041

>Thx, Chris

More information about the questions mailing list