[ntp:questions] Re: A theoretical question

Eric ejensen at spamcop.net
Fri Nov 7 21:13:01 UTC 2003


They all work.  

I vote for (c). This allows the local machine to be accurately synced
when the link is up, and still serve time (from the local clock,
fudged to stratum 10) even when disconnected, which will at least keep
all machines on the lan in sync with the local NTP server.  This is
what most dial-up folks do.  

- Eric

On 07 Nov 2003 01:35:58 +0500, Denis Zaitsev <zzz at natasha.ward.six>
wrote for the entire planet to see:

>There is a small LAN with one Linux machine has a dialup access to the
>internet.  And this machine is also a NTP server for the LAN.  It uses
>the LOCAL (127.127.1.0) clock.  I want to synchronize this machine
>with an external NTP servers thru the PPP sessions.
>
>a) I do this that way: when the PPP link is up, I run ntpdc and
>unconfig 127.127.1.0 and config some quantity of the external servers.
>When PPP link is down, I do the reverse work.  And it seems that this
>method works.
>
>b) There is another way: to config the external servers once and do
>not do any swaps.  The NTP server lives happily without the LOCAL
>clock, and just complains sometimes when PPP is off about an
>unreachability of the external servers.  When PPP is on my NTP server
>reestablishes the connections and resets the time, if it is a need.
>And again, it seems that it works.
>
>c) And there is the third way: (b) plus LOCAL clock at low priority,
>to be used at the absence of an access to the "real" servers.
>
>So, my question is: what are the weakness of each of these methods?
>They all seem to work, but I have a paranoid feeling, thet they give
>the different results...
>
>Thanks in advance.




More information about the questions mailing list