Hal Murray wrote:
That's what I thought at first, but looking at my clients stats it
looks like roughly 2/3 of the clients are SNTP clients. If I'm not
mistaken, these clients will query the DNS, pick up one server, do
one NTP query and be done with it. 
    

How do you tell if a client is running some SNTP package?

I'd expect a cron job running ntpdate would also do a DNS lookup and then 
make a single probe.
  

That is the same thing, isn't it?  Other than that ntpdate was written by our expected
ntpd author, and includes some functionality that maybe not all SNTP implementations
have, like using adjtime for making small adjustments, using ntpdate from a cronjob
is equivalent to using many SNTP packages.

This type of usage has advantages and disadvantages.   doing a DNS lookup for each
request, and making maybe one request an hour, at least "guarantees" that you will
get a responding server once in a while.
When you use ntpd and one or more servers in /etc/ntp.conf, it can happen that the
servers that were selected at boottime have ceased to exist or no longer want to be
a timeserver and you remain polling them for the entire uptime of the system (which
may be months or years).  So, monitoring and manual action is required.

I think when configured with a domainname (not a fixed address), ntpd should do a
DNS lookup once in a while.  Or at least do it when it detects the server is dead or
not locked.   But probably this list is not the place to get such a request communicated.

Rob
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to