NTP: Driving Me Nuts

[David Marshall]

I'm having an awful time trying to get NTP working on some new servers.

Consider three machines:

jeffy: my home machine, runs 5.4 with a very sleek kernel, sits behind a router.

web1: one of the new servers, running 5.4-p2 with a kernel
configuration that *only* has options INET6 commented out:

root at web1# diff GENERIC WEB
25c25
< ident         GENERIC
---
> ident         WEB
32c32
< options       INET6                   # IPv6 communications protocols
---
> #options      INET6                   # IPv6 communications protocols


web2: another of the new servers, running 5.4 with GENERIC

All three have the same /etc/ntp.conf:

server sundial.columbia.edu
driftfile /var/db/ntp.drift

All, of course, have ntpd_enable="YES" in /etc/rc.conf

Here are the relevant lines from the log when I run /etc/rc.d/ntpd
start, after making sure it is stopped, of course.  When any of them
is stopped "netstat -n | fgrep 123" yields no lines.

jeffy:

Jun 11 23:24:53 jeffy ntpd[90141]: ntpd 4.2.0-a Mon May  9 15:42:44 PDT 2005 (1)
Jun 11 23:24:53 jeffy ntpd[90141]: no IPv6 interfaces found

web1:

Jun 12 02:28:23 web1 ntpd[783]: ntpd 4.2.0-a Sun Jun 12 00:46:05 EDT 2005 (1)
Jun 12 02:28:23 web1 ntpd[783]: no IPv6 interfaces found
Jun 12 02:28:23 web1 ntpd[783]: bind() fd 6, family 2, port 123, addr
0.0.0.0, in_classd=0 flags=8 fails: Address already in use

web2:

Jun 12 02:24:28 web2 ntpd[32792]: ntpd 4.2.0-a Sun May  8 06:01:21 UTC 2005 (1)
Jun 12 02:24:28 web2 ntpd[32792]: bind() fd 9, family 2, port 123,
addr 0.0.0.0, in_classd=0 flags=8 fails: Address already in use


'ntpq -p' on jeffy returns a normal looking ntpq result.  'ntpq -p' on
either web1 or web2 eventually times out.  However, if I put a
"restrict" into their ntp.conf I get something like

     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 hickory.cc.colu .INIT.          16 u    -   64    0    0.000    0.000 4000.00

that never changes, even after several hours.

I've read a lot of similar problem reports, but none of them ever seem
to have a definitive answer.

Can anyone help?  I'm really mystified.

The only thing I have left to try is that I have noticed that jeffy
has NO_INET6 = true set in /etc/make.conf, whereas web1 does not have
this.