[ntp:questions] Broadcast server with external one

Glencu Glencu glencek at gmail.com
Wed Jul 11 12:46:07 UTC 2012


Thanks Dave,


I forgot to mention that in first case I'm killing right after the sys peer
is found.

Just want to make sure here:
If broadcast client receives packet with root dispersion greater than
MAXDIST, than it will remove that server from sys peer position - it that
correct?


Regards

2012/7/10 Dave Hart <hart at ntp.org>

> On Tue, Jul 10, 2012 at 08:58 UTC, Glencu Glencu wrote:
> > There is a PC which acts as broadcast server for my subnet AND as unicast
> > client which points to external server. Everything works ok and my
> question
> > is rather theoretical. So, I'm killing the connection with external
> server
> > and I'm getting two results:
> >
> > 1. If external server is working for some single minutes ( like 20
> > minutes)- I synchronize PC and then I kill the connection the brodcast
> > server stops to send the packets
> >
> > 2. If external server is working for longer period of time (like two
> days)
> > - I synchronize PC and then I kill the connection the broadcast server
> does
> > not stop to send packets - it even does not do that after one or even two
> > days.
> >
> > I would expect that in second case server would also stop to broadcast.
> Am
> > I right? I think it has something common with external server stabuility.
> > Sorry, If the topic was already covered. I couldn't find it anywhere.
>
> I don't understand the difference you're seeing between the two cases
> unless the broadcaster has never synchronized before the external
> server becomes unreachable.  You can tell it's never been synchronized
> if "ntpq -c rv" output includes "leap=11".
>
> In the second case the behavior you're seeing is correct.  The
> broadcaster's root dispersion will increase so long as it is unable to
> reach its source(s).  It's up to the broadcaster's clients to decide
> how much root dispersion is too much and rule it out.  The default for
> ntpd is 1.5 seconds and can be customized with "tos maxdist":
>
> http://www.eecis.udel.edu/~mills/ntp/html/miscopt.html#tos
>
> Cheers,
> Dave Hart
>


More information about the questions mailing list