[ntp:questions] Troubleshooting who's at fault

David Woolley david at ex.djwhome.demon.co.uk.invalid
Thu Jun 25 22:13:14 UTC 2009


Ronaldo Mexico wrote:

> 
> Unfortunately, it's the RHEL server (147.159.120.206).  The PTS is on
> 147.159.120.131.  There are no other NTP servers in the private
> network.

This is why configuring a local clock driver is dangerous. Especially as 
you are mixing in w32time, that won't understand orphan mode (which 
needs very recent versions of ntpd, anyway).

Either your real reference clock is broken, or you are trying to operate 
an isolated time islands (i.e. all machines the same time, but time 
progressing at an arbitrary different rate from real time).  If orphan 
mode is not an option, you need to be very careful about how you 
configure your local clocks.

Something else that is worrying is that your are getting stratum 2, but 
a server loop, with other side of the loop being at stratum 10.  That 
suggests you are running a broken ntpd server.  The two sorts of broken 
server that I am aware of are non-Windows 2003 versions of w32time, 
which and older versions of openntpd, both of which are actually SNTP 
servers, and both of which incorrectly report stratum 2, regardless of 
the input stratum.  I'm not sure if the recent versions of w32time have 
this bug by default.  With the latest versions you need to change quite 
a few registry entries before they comply with the protocol.

Precision -20 is too good for w32time, so possibly openntpd.

Maybe it simply has the local clock fudged to stratum 1; ntpd will, 
normally disregard a local clock if there is a valid clock, but the loop 
and the lack of enough non-locals to outvote, maybe frustrates that here.

Life will be a lot easier if you install a radio clock and delete the 
local clock driver lines from all your machines.




More information about the questions mailing list