[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ntp dies after a while....

On Sep 30, 2009, at 4:15 AM, Niels S. Eliasen wrote:

hi guys
Noticed that after some days(even weeks) ntpd dies for no apparent reason......

Anyone seen this ?
and are there any known fixes ?

from daemon.log I see:

Sep 28 12:49:59 munin ntpd[19926]: synchronized to, stratum 2 Sep 28 12:49:59 munin ntpd[19926]: time correction of 2295 seconds exceeds sanity limit (1000); set clock manually to the correct UTC time.

but as the daemon is running... surely it should correct this before getting into this "exceeds sanity limit" issue ??

I don't know how the system clock gets so far off from UTC -- maybe indicative of a hardware problem? Are you, maybe, using anything besides NTP to set your clock? Maybe one of your servers ( springs to mind) is serving bogus time?

In any case, the "step the clock if necessary -- no matter how far off it may be" ( "-g" option to ntpd) mode is a one-shot to take care of machines that have lousy (or non-existent) on-board CMOS clocks. They come up after a reboot with very-wrong or simply random time. After the time is judged to be in sync, ntp refuses to step the clock if it's discovered to be outside of the sanity window.

The logic of this policy is that if the clock is synced, and it suddenly goes out of sync, there is something terribly wrong, and the sysadmin needs to get it fixed. Silently stepping the clock by more than 1000 seconds is not the right thing to do. It could cause much more harm than good. So the ntpd daemon commits suicide rather than contribute to a larger problem.

This has been discussed endlessly on the NTP mailing lists. It's a decision made by Dave Mills himself, and he refuses to change his mind.

For details, see "man ntpd"


Reply to: