[ntp:questions] NTP settings for machine with irregular, short connections to the Net

Maarten Wiltink maarten at kittensandcats.net
Tue Sep 11 12:05:50 UTC 2007


"Rikishi 42" <fsck_spam at telenet.be> wrote in message
news:65ugr4-b9v.ln1 at whisper.very.softly...

> [The laptop] brings a 'correct time' from the outside.

I wonder about that. On my own machines, I often see significant steps
(up to several seconds) when rebooting them. It's worth testing at
least.

> Only concern is: when the machines see the laptop, can they sync realy,
> really fast, so that their NTP's work is done in a very short lapse of
> time? In that case at least, the clocks would be within the same second
> most of the time.

If you designate one of the isolated systems the local server, you
could have it check for the laptop's presence every minute or so, and
restart the NTP service if it comes on-line. This assumes that starting
the NTP service involves an initial step if necessary. Not very friendly,
but workable.

A better option might be to estimate its offset, dissipate that, and
return to its old (almost correct) notion of one second per second.
I don't know if NTP can handle this. I can imagine that if an
association comes back on-line for long enough to estimate the offset,
then goes away before it's reduced to near zero, the frequency correction
remains in effect indefinitely.

One thing I've done is disable the initial step, adjust the drift file,
start NTP and let it run for awhile, then revert the drift file and start
NTP again. And that might be automated. (It also caused the usual cries
from here that this wasn't what NTP was designed for.)


>> This is why we have shell scripts.
>
> The running of which is subject to old people's forgetfullness.
> (and I'm counting myself with those :-)

Shell scripts can be made to run whenever network interfaces are
brought up.

Groetjes,
Maarten Wiltink





More information about the questions mailing list