[ntp:questions] ntpdate removal is coming

Steve Kostecke kostecke at ntp.org
Sat Dec 10 18:23:23 UTC 2011


On 2011-12-10, Harlan Stenn <stenn at ntp.org> wrote:

> Steve wrote:
>
>> Bruce wrote:
>>
>> > Those times apply *after* enough samples have been gathered for
>> > an offset estimate, which happens *after* a system peer has been
>> > selected. That can take many minutes. With ntp-wait, the boot
>> > sequence would be effectively stalled for the duration.
>>
>> There is no need to use ntp-wait.
>>
>> Use sntp.
>>
>> Use a separate 'ntpd -gq' invocation.
>
> Why?

Because the current practice of using ntpdate to initially set the clock
is GoodEnough(tm) for most users; and both sntp and 'ntpq -gq' are, for
the most part, the functional equivalent of ntpdate.

> ntp-wait will block until leap != 11, which means that ntpd is
> satisfied that it has a sufficient understanding of time sync that
> steps should not be needed and timekeeping is on its way to stability.
>
> ntp-wait can be used to wait until this is true, at which time it is
> considered safe to start things like database servers and dovecot.
>
> If you are not running any of these sort of services there is no need
> to run ntp-wait.

It may be interesting to find out how many people currently use
ntp-wait.

>> > 2. your estimate of "fully sync'd in 11 seconds' time" seemed
>> > overly optimistic (and still does, even with minimum polling times and
>> > maximum slew rate).
>> 
>> The 11 second figure applies to the initial setting of the clock with
>> 'ntpd -gq' (which is analogous to the use of ntpdate for the same
>> purpose) and not to disciplining the clock to maximum stability.
>
> I'm not gonna dig deeply into that one, but I will say that I remember
> seeing similar values for "delay" when using ntp-wait.

The 11 second figure for 'ntpd -gq' came from warm-restart tests I ran
on my previous home LAN under optimal conditions (local DNS and a LAN
time server).

-- 
Steve Kostecke <kostecke at ntp.org>
NTP Public Services Project - http://support.ntp.org/



More information about the questions mailing list