[no subject]

> > Another question is whether it's correct to say "System clock". The
> > messages don't actually print the offset of the system clock, but
> > rather the change in the offset relative to the NTP time that chronyd
> > is tracking. On start that is the same as the offset of the system
> > clock, but if there is an unfinished correction, it's the difference
> > between the old offset and new offset.
> 
> I'm sorry, I don't understand.  What does that mean, "the change in the offset"?
> 
> The quantity ( system_clock_time - NTP_server_time ), I think I understand.
> 
> Now, are you describing something like
>  (( new_system_clock_time - new_NTP_server_time ) -
>  ( old_system_clock_time - old_NTP_server_time ))?
> 
> Are you saying that there is a system log message that would be saying "the system clock was behind by 5 seconds, but now, after some correction, the system clock is behind by only 3 seconds, and the offset has been 'reduced' by 2 seconds"?

No, I think it's something else. The issue is that the clock is
normally corrected by slewing (instead of stepping) and that takes
time. Before a correction is finished, another measurement can be made
and the correction needs to be adjusted. That adjustment is what
chronyd reports with the "clock wrong by" message.

Normally, if both the time source and system clock are stable, that
adjustment should be small and we don't expect to see a second "clock
wrong by" message. But if one of them is not stable enough (we
generally don't know which), the adjustment can exceed the logchange
threshold and another "wrong by" message will be logged.

For example, if the initial offset was 5 seconds and the system clock
was already corrected by 2 seconds when another measurement is made,
which says the offset of the system clock is now actually 4 seconds
instead of the expected 3, a message will be logged that the clock is
wrong by 1 second. The question is which offset of the system clock
should be reported (4 or 1) and when it should be reported (abs(4) >
logchange or abs(1) > logchange).

> > There are at least the following options:
> > 1. change the message to better describe what is behind or ahead
> > 2. change the message to print the real offset of the system clock
> >    a) print it on each update as long as the offset is larger than
> >       logchange
> >    b) print it only when the change in the offset is larger than
> >       logchange
> > 
> > Thoughts?
> 
> If I am understanding, I think I prefer option '2.a)'.

What I don't like about 2.a is that if there is a large initial offset
and stepping is not enabled by "makestep", there may be a large number
of the "clock wrong by" messages in the log, even if everything is
working as expected.

-- 
Miroslav Lichvar

-- 
To unsubscribe email chrony-dev-request@xxxxxxxxxxxxxxxxxxxx with "unsubscribe" in the subject.
For help email chrony-dev-request@xxxxxxxxxxxxxxxxxxxx with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.