[ntp:questions] NTP time synchronization issue in linux embedded target

Harlan Stenn stenn at ntp.org
Wed Sep 3 21:19:04 UTC 2014

Murugesh S writes:
> Hello,
> This email is regarding NTP time synchronization issue that we see in
> our linux embedded systems.  NTP version that is being used: 4.2.6p4.
> Brief description about issue: Our linux system is configured with NTP
> running 4.2.6.p4. After couple of weeks the system clock has local
> time ahead /faster than actual time in NTP.  The time delta is between
> 15 minutes to 30 minutes. The ntp.drift file reads value 500.00. This
> issue is observed in more than one system.

That means that NTPD has decided your system clock needs more than a
500ppm correction to keep the clock sync'd.

That is the largest frequency correction NTP can handle, and that points
to a frequency problem with your clock.

If this is a "short-term local clock insanity" issue, you might be able
to fix it by removing the drift file and restarting ntpd.

If this is a long-term real problem with the onboard clock, then if your
kernel supports the "tick" variable, you will need to change the "tick"
value in your kernel to have your base system clock keep better time.

The tickadj script should be able to help with this.

Basically, you need to calculate the drift correction and apply that to
yout tick value.

If your local clock is running fast by 500ppm, instead of using a value
of, say, 10000 for the tick, you might need a value closer to 9995.

If your local clock is actually running fast by 1,000ppm, your tick
value will need to be closer to 9990.


More information about the questions mailing list