[ntp:questions] Server down but time staying in sync

raju dantuluri dsraju0 at gmail.com
Tue Jul 14 18:26:32 UTC 2009


We have observed that the NTP client never appear to be out of sync
even though the NTP server is down! It's important for us to know that
the server is down/not available!
we use ntpq to know about its association! Ideally we should use more
NTP servers but we use only one at this time!

Steps to reproduce:
--The client node (10.1.1.203) is configured and made sure that it is
in sync with the server:
$ cat /etc/ntp.conf
restrict default ignore
restrict 127.0.0.1
server 172.16.1.49 iburst
restrict 172.16.1.49
$ ntpq -np
     remote           refid      st t when poll reach   delay
offset  jitter
==============================================================================
*172.16.1.49     209.123.234.24   3 u  703 1024  377    0.514
-3.268   1.617
$

--Later I shutdown the NTP server on 172.16.1.49

Now the "when" field keep increasing as there is no response from the
server.
here is the ntpq output at different times!

$ ntpq -np
     remote           refid      st t when poll reach   delay
offset  jitter
==============================================================================
*172.16.1.49     209.123.234.24   3 u  740 1024  377    0.514
-3.268   1.617
$ ntpq -np
     remote           refid      st t when poll reach   delay
offset  jitter
==============================================================================
*172.16.1.49     209.123.234.24   3 u  53m 1024  370    0.514
-3.268   1.617
$ ntpq -np
     remote           refid      st t when poll reach   delay
offset  jitter
==============================================================================
*172.16.1.49     209.123.234.24   3 u  63m 1024  370    0.514
-3.268   1.617
$ ntpq -np
     remote           refid      st t when poll reach   delay
offset  jitter
==============================================================================
*172.16.1.49     209.123.234.24   3 u   9h 1024    0    0.514
-3.268   0.000
$ ntpq -np
     remote           refid      st t when poll reach   delay
offset  jitter
==============================================================================
*172.16.1.49     209.123.234.24   3 u  10h 1024    0    0.514
-3.268   0.000
$ cat ntp.drift
89.969
$ ntpd --version
ntpd - NTP daemon program - Ver. 4.2.4p3

This node runs on the Linux 2.6.25.14 #1 SMP kernel!

Observe that the "reach" is zero but the tally code "*" indicates that
it is synchronized!
In my previous attempt I've observed that "when" went up to few days
before reboot while the time staying in sync!
I could restart ntpd that will recover from this state but this is not
what I am looking for!

How do I estimate the time before the tally code becomes " " (space)!




More information about the questions mailing list