[ntp:questions] Troubleshooting who's at fault

Dave Hart davehart at gmail.com
Fri Jun 26 09:00:57 UTC 2009


On Fri, Jun 26, 2009 at 7:04 AM, David Woolley wrote:
> Harlan Stenn wrote:
>
>> If you are using LOCAL as a fallback on your client, and your upstream
>> server is using LOCAL ias the name for its PTS-sync'd refid, then the client
>> just sees that the 2 sources it knows about are using the same refid and
>> will flag that as a loop.
>>
> That doesn't sound sensible.  It implies you cannot peer two machines
> that have the same reference clock technology.  Is the problem that
> reference clocks, as a special case, use the same reference ID as the
> stratum 1 server that uses them, and, by the time the check is done, the
> distinction between a reference clock and a server has been lost?

I think another part of the equation that causes the loop detection to
fire here is the reference clock has been fudged to a numerically
higher stratum.  With the refclock left at the default stratum 0 (and
so the ntpd attached at stratum 1) the loop detection will not fire
for two peered ntpds using the same driver and unit number.  In other
words, I believe the loop detection treats stratum 1 refid's
specially.

Cheers,
Dave Hart



More information about the questions mailing list