[ntp:questions] Local clock - sync issue

David Woolley david at ex.djwhome.demon.invalid
Mon Nov 8 23:30:46 UTC 2010


Stephen Vaughan wrote:
> Connectivity is fine to the ntp servers, if I restart ntpd it starts syncing with an external clock immediately. This article could be the fix:
> 
> http://www.novell.com/coolsolutions/feature/15345.html

This advises some very anti-social practice, namely using burst on 
public servers.  This is likely to get you banned from those servers. 
Also, it doesn't explain why you have zero reachability.

Getting the detailed diagnostics, by using rv on the association 
numbers, may give a better clue.

I do have some concern that a local clock could result in all the others 
being false tickers, because the error band on the local clock is too 
small, but ntpd is supposed to discriminate against the local clock.  In 
any case, false tickers still have a non-zero reachability.

Actually, one possibility, is that the error statistics on the remote 
servers are wrong, resulting in their confidence intervals not 
overlapping.  That doesn't really work for the Novell example, as they 
only have one server.  Unfortunately, they've used ntpdc peers, rather 
than nptq peers, so one can't really tell how the servers are being treated.

However, in all cases, the basic problem is using a local clock driver 
for no valid reason.




More information about the questions mailing list