time sync with ntp not working correctly?
John DeDourek
dedourek at unb.ca
Tue Feb 1 13:02:00 UTC 2005
Note the column called "reach". On the server it has 377.
That's the octal (base 8) representation of an 8-bit quantity.
(377 = 11111111 binary). Periodically a machine running ntp
sends a time query to its configured servers. When it gets a
reply it shifts a 1 into the right of "reach"; no reply, it
shifts a 0. Thus on the server, the last eight queries to
its servers got a reply: 11111111. On the clients 00000000;
that's obviously bad. So the clients are not getting responses
to their queries.
Two usual problems:
-- Either the queries or the responses to the clients are not
getting through the network; most commonly firewall issues;
occasionally network problems like routing; check the latter
by ping. The former requires an investigation of the firewall
setup
-- Bad ntp configuration on the server (refusing to accept queries)
or on the clients (refusing to accept responses). You
unfortunately didn't shows us the ntp configuration on the two
machines. Most common problem is that the meaning of "restrict
notrust" changed between older and newer versions of ntp. If
you are reading old guidlines, or have upgraded to the newer
ntp and used the configuration files from the old one, that
could be the problem. Have a quick look at /etc/ntp.conf;
if the word "notrust" appears on any "restrict" lines, try
editing it out (saving a copy of the old configuration first).
If you just copied an old configuration file over the new one
after upgrading (and happened to save the original installed
configuration, which I highly recommend), a good procedure would
be to go back to the ntp.conf installed by the upgrade and edit
your own server lines in (making changes as appropriate). In
any case, we would need to see the configuration files to
comment further.
Götz Reinicke wrote:
> Hi,
>
> today I checked the time on some servers and found that they differ by a
> couple off minutes. Without teh ntpd running a "ntpdate gaugin" syncs
> the clock.
>
> All servers are configured using one timeserver using ntp. The clients use
>
> server 172.17.1.251
>
> The Relay ntp server uses public servers.
>
> I've reed the "http://www.tldp.org/HOWTO/TimePrecision-HOWTO/ntp.html"
> and "ntpq -p" on two clints is:
>
> remote refid st t when poll reach delay offset jitter
> ==============================================================================
>
> gaugin.filmakad .INIT. 16 u - 128 0 0.000 0.000
> 4000.00
>
> on one client:
>
> remote refid st t when poll reach delay offset jitter
> ==============================================================================
>
> gaugin.filmakad 0.0.0.0 16 u - 1024 0 0.000 0.000
> 4000.00
>
> The server:
>
> remote refid st t when poll reach delay offset jitter
> ==============================================================================
>
> LOCAL(0) LOCAL(0) 10 l 55 64 377 0.000 0.000
> 0.008
> +nic1.belwue.de ntp1.belwue.de 2 u 697 1024 377 2.687 0.518
> 0.343
> +news.belwue.de nz11.rz.uni-kar 2 u 738 1024 377 2.374 -0.747
> 0.188
> *dns4.belwue.de hora.cs.tu-berl 2 u 697 1024 377 2.173 -0.530
> 0.205
>
> What could be wrong?
>
> Regards
>
> Götz Reinicke
More information about the fedora-list
mailing list