On Mon, 14 May 2007, Stuart Henderson wrote:
> On 2007/05/13 23:06, John Nietzsche wrote:
> > */5 * * * * /usr/sbin/rdate -4ncva
>
> -c corrects for leap seconds
>
> > */5 * * * * /usr/sbin/rdate -4cva gw |
>
> and here you do it again i.e. you are correcting time coming from a
> source which is already corrected.
>
> I would really recommend against using rdate like this, it jumps the
> clock. ntpd skews the clock (makes it run slightly fast or slow until
> the time is correct), so you don't miss out on any seconds (which
> sometimes skips cron jobs, makes logging more confusing, and can
> cause a lot of trouble with some other applications).
While I agree with the advise, this is not true when the -a flag is
given to rdate, in that case rdate uses adjtime(2).
ntpd is of course much nicer, since its adjust the clock frequency as
well, and poses a very light stress on the server: once time is
synced, queries do not happen a lot. Not to speak of the ability to
use time sensors and multiple time sources to provide redundancy.
-Otto
