[ntp:questions] Tool to sample/monitor an NTP server?

PhilipPeake philip.peake at gmail.com
Mon Nov 9 23:15:08 UTC 2009


My problem is a client with several systems using a single GPS-based
NTP server.
A couple of these systems have had occasions where the clock has
suddenly stepped around 10,000 seconds backwards:

Oct 15 13:22:03 xxx xntpd[403]: [ID 261039 daemon.error] time error
-10369.999603 is way too large (set clock manually)
Oct 15 10:29:18 xxx ntpdate[12547]: [ID 774510 daemon.notice] step
time server 10.215.247.73 offset -10370.000010 sec
Oct 15 10:29:18 xxx xntpd[12549]: [ID 702911 daemon.notice] xntpd 3-
5.93e+sun 03/08/29 16:23:05 (1.4)
Oct 15 10:29:18 xxx xntpd[12549]: [ID 301315 daemon.notice] tickadj=
5, tick = 10000, tvu_maxslew = 495, est. hz = 100
Oct 15 10:29:18 xxx xntpd[12549]: [ID 266339 daemon.notice] using
kernel phase-lock loop 0041, drift correction 0.00000
Oct 15 10:29:18 xxx xntpd[12549]: [ID 266339 daemon.notice] using
kernel phase-lock loop 0041, drift correction 159.93300

Shortly afterwards, the time steps again back to the correct time.

I suspect that the NTP server is having issues, especially since this
same thing has happened on more than one server. The events are not
synchronised, so I suspect a short duration time glitch only seen if
the NTP client happens to poll at the "right" time.

What I want is a client to poll the NTP server (say once per second)
and log the time received. A simple perl script can then look for any
large jumps in the recorded time. This will tell me if the NTP server
is really at fault.

I am having problems finding a tool to do this. I suspect that
something of the sort must already exist somewhere ... so if anyone
has any suggestions, please do let me know.




More information about the questions mailing list