Inconsistency between ntpd and ntpdate

Question

I am trying to synchronize two computer in my local network. To do so I use NTP on one computer to synchronize with an ntp pool. Then I use ntp on the second computer and I tell it to synchronize with my first computer in the /etc/ntp.conf

server first_computer iburst

now when I do ntpq -p on the second computer I get:

     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 scitos.magazino 85.25.105.105    2 u   45   64    0    0.000    0.000   0.000
-news02.nierle.c 192.53.103.104   2 u   63   64  377   29.065   -2.921   0.623
*ns1.customer-re 40.33.41.76      2 u   63   64  377   18.636   -0.088  22.543
+stratum2-2.ntp. 129.70.130.71    2 u   61   64  377   30.640   -0.063   0.839
+ntp1.wtnet.de   36.224.68.195    2 u   57   64  377   34.993   -1.956   0.294

where scitos is my first computer. Everything seems nice. But when I do ntpdate -q scitos

I get:

server 10.8.64.62, stratum 2, offset 0.040108, delay 0.02666
 5 Aug 13:42:54 ntpdate[6928]: no server suitable for synchronization found

Why is there still a delay for ntpdate?

Answer 1

The delay does not mean a delay between your computer and the source. Rather, this is the time taken for a roundtrip of the ntp message.

There is a fundamental uncertainty as to how to synchronize two clocks when messages between them take some amount of time, greater than zero. NTP tries to do the best guess it can. However, if for example the network messages take a drastically different time in each direction, it will not be accurate.

In other words, you shouldn't worry about the delay figure. It is a statistic provided for people tweaking NTP and does not represent the level of discrepancy.

See for example http://tech.kulish.com/2007/10/30/ntp-ntpq-output-explained/