[ntp:questions] Re: ntp.conf For Clients

David Woolley david at djwhome.demon.co.uk
Tue Jun 21 20:15:41 UTC 2005


In article <mailman.104.1119362028.91305.questions at lists.ntp.isc.org>,
Stewart Pelegan <spelegan at csc.com> wrote:

> Right now, I've got the following in the clients' ntp.conf:

> broadcastclient
> multicastclient 224.0.1.1

It's not very sensible to have both.  You should get things working
properly without multi/broad-cast before you add this complication.

> enable auth monitor

I'm not an expert on auth, but you should not have it, initially,
when debugging without broadcast.   I think it may be needed for
broadcast modes.  I cannot see any indication of the key files
that it must need.

> and the servers have:

> server 127.127.5.0 prefer
> server 10.20.30.41
> server 127.127.1.0
> fudge 10.20.30.41 stratum 1

The stratum 1 here will be ineffective.

> fudge 127.127.1.0 stratum 1

Setting a local clock to stratum one is not something you want to
do lightly.  I can see no valid reason for doing it here.  I am pretty
sure that having this local clock will make the configuration unstable
when the radio clocks fail.  The only time to have one at stratum one
is if the local clock is being disciplined by something other than ntpd,
and even then there is the problem that ntpd will not know if that
source has failed.

> broadcast 224.0.1.1 ttl 6

ttl 6 is too large if all the clients are on the same sub-net.  Simple
broadcast would be enough.

> This is a PRIVATE message. If you are not the intended recipient, please 

Nothing posted to comp.protocols.time.ntp is PRIVATE, and therefore
nothing posted to the feeder email gateway is PRIVATE.  However,
the gateway is broken, so, if possible, please use the newsgroup
directly.

I am assuming that the legal notice is bogus, at least as to confidentiality
and I can ignore it and copy into my reply.



More information about the questions mailing list