Rob Janssen wrote on 10-2-2008 13:21: > Udo van den Heuvel wrote: >> Hello, >> >> I got one of them emails even though the machine is running OK; >> I did not change anything about ntp, the box has been up for 71 days and >> yet the graph at pool.ntp.org shows an offset going downhill. >> Also the network config has not been changed. There is slight traffic >> shaping, as has been there for ages on my side of the connection, just >> to prevent issues like these. >> >> What is the most likely cause? >> > In my experience (at least for my own system) the offset graph is > showing the network latency towards the monitoring server, rather than > the quality of my clock. > I see no irregularities over the past days, but in the past the offset > sometimes was quite high. > > I keep a couple of external reference clocks in addition to my two local > clocks (GPS and DCF), so I can monitor my own ntp's view on the offset > and compare it to what the monitoring server thinks. > When your own "ntpq -p" does not show an offset and there is one at > www.pool.ntp.org, it must be somewhere in the network.
That's correct, it is a network issue. The monitoring system is in California, for us here in europe there's an ocean and a continent in between. I did some simple tests on this a few months ago; my conclusion was that the monitoring system can be used to see if my server is reachable, nothing more. The inaccuracies introduced by the network are in the order of ten to fifty ms, with peaks far over 100 ms, while a carefully run stratum 2 server doesn't have to be more than a millisecond off. Note that the pool monitoring system gives only a small penalty for this, it's unlikely that you'll be thrown out off the pool because of this issue alone. But you'll have to find another way to judge your servers accuracy. regards, Jan Hoevers _______________________________________________ timekeepers mailing list [email protected] https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
