[ntp:questions] Stratum goes back to 16 all the time

david Woolley david at djwhome.demon.co.uk
Sun Jan 14 10:32:13 UTC 2007

In article <1168766937.300189.203640 at v45g2000cwv.googlegroups.com>,
independence at blinkenlights.se wrote:

> -nissan.ifm.liu. .PPS.            1 u   12   64    3   28.275  -259.30 15.966

You have either not been running ntpd long enough (less than a minute,
given that you are using iburst) or you have had a recent time step,
as you have only had 2 or three polls since the last time there was a
total reset of the clock filters.  The offset has already exceeded the
limit at which there will be another step.  Your clock is fast, which
excludes the common lost interrupts cause, so one of the following is
likely to be the case:

- the clock frequency on your PC is more than 500ppm fast - replace the
- the effective clock frequency is varying as the result of power management -
  there are reports that there are boot time parameters that may cause simpler 
  timing mechanisms to be used (note this is a kernel issue, not an NTP one);
- there is some other software that is trying to set the clock (e.g. you are
  running hwclock in a cron job) - remove or disable it;
- you have a network connection that occassionally has very high assymmetric
  traffic loads - investigate the huff and puff option in ntpd.

Having syslog entries showing the time steps might help to exclude some of
these, e.g. asymmetric delays tend to cause almost equal positive and 
negative steps, whereas an out of range clock frequency will give consistent
steps in the same direction and consistent intervals.

More information about the questions mailing list