[ntp:questions] Re: Correcting my time servers clock drift on Alpha ES40s / Tru64

Tom Smith smith at cag.zko.hp.com
Fri Oct 21 04:49:29 UTC 2005


By the way...

> brett.sander at tenix.com wrote:
 >>
>> 20 Jul 17:40:44 xntpd[764]: frequency initialized -5.490 from
>> /cluster/members/{memb}/etc/ntp.drift
 >>

If my arithmetic is right, this is a negative adjustment of about
200 seconds over 14 months ((5.490 X 60 X 60 X 24 X 30 X 14)/1000000).

>> phillip:
>>
>> 20 Jul 17:41:50 xntpd[764]: frequency initialized -10.384 from
>> /cluster/members/{memb}/etc/ntp.drift

This is a negative adjustment of about 375 seconds over 14 months.

 >> Currently they have been up and running for about 14 months and the
 >> clocks are now about 30 minutes slow.

This is about 36 parts per million, if the clocks were, in fact, correct
14 months ago. So your hardware and OS may be keeping perfectly acceptable
time all by itself, possibly slow within specified tolerances, but because
you have no reference to reality and the clocks are just floating and left
to their own devices, NTP has decided (by this date in July) to further
slow your clocks by about 0.47 and 0.89 seconds per day (-5.490 and
-10.384 parts per million), respectively.

Unless you can manage to find real reference clocks, including, for
example, NTP servers that might sit on your firewall, if any, you may be
better off with no drift file and with the local clock configured
on only one of your systems. Even then, you should expect to gain or
lose time relative to reality.

 >> terrance's ntp.conf:

 >> peer phillip
 >> peer 127.127.1.1

 >> phillip's ntp.conf:

 >> peer terrance
 >> peer 127.127.1.1

In any case, the local clock should be declared as a "server" and not a "peer".

-Tom




More information about the questions mailing list