Problem with xntp after reinstall on openSUSE 10.3

Problem with xntp after reinstall on openSUSE 10.3

Post by Marc Muehl » Wed, 17 Oct 2007 00:41:05


Hello,

I run a xntp server on Suse 10.0 for the last 2 years. Last week I reinstalled
the server using openSUSE 10.3. I setup the server to run chroot (in
/etc/sysconfig/ntp) like before and use the same ntp.conf. But now the time is
allways wrong after some hours.

Meanwhile I took all restrict rules out of the file (but I had them in before
too and it worked for a long time), but still have this problem. This is the
current ntp.conf I use:


server 127.127.1.0 # Local clock
fudge 127.127.1.0 stratum 10 # Local clock is unsynchronized (stratum 10)

server 0.de.pool.ntp.org iburst prefer
server 1.de.pool.ntp.org iburst
server 2.de.pool.ntp.org iburst

driftfile /var/lib/ntp/drift/ntp.drift
logfile /var/log/ntp


About two minutes after starting xntpd:
# ntpq -p
remote refid st t when poll reach delay offset jitter
==============================================================================
LOCAL(0) .LOCL. 10 l 21 64 1 0.000 0.000 3.906
*213.ip-net02.em 192.53.103.108 2 u 4 64 1 16.846 113.006 48.404
+www.alter-provi 217.91.44.17 2 u 3 64 1 27.271 147.323 39.202
+berlin.zq1.de 194.25.115.122 2 u 2 64 1 28.322 121.759 46.183


About 5 minutes after startup it fall back to local and ntpd shows:
# ntpq -p
remote refid st t when poll reach delay offset jitter
==============================================================================
*LOCAL(0) .LOCL. 10 l 17 64 37 0.000 0.000 3.906
213.ip-net02.em 192.53.103.108 2 u 16 64 37 16.435 1247.22 958.049
www.alter-provi 217.91.44.17 2 u 14 64 37 28.379 1298.52 968.629
berlin.zq1.de 194.25.115.122 2 u 12 64 37 27.911 1269.07 973.514


The ntp logfile just says:
# cat /var/log/ntp
15 Oct 17:31:09 ntpd[6534]: synchronized to 62.52.175.213, stratum 2
15 Oct 17:31:09 ntpd[6534]: kernel time sync status change 0001
15 Oct 17:35:11 ntpd[6534]: synchronized to 85.25.132.119, stratum 2
15 Oct 17:35:15 ntpd[6534]: synchronized to LOCAL(0), stratum 10


After 2 days my ntp server has allready a delay to the real time of about 20
minutes. Any idea why it allways fall back to local? The internet connection
is allways available.


Regards
Marc
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Richard B. » Wed, 17 Oct 2007 01:54:32

arc Muehlfeld wrote:

My first guess would be that your local clock is off by more than 500
Parts Per Million (PPM). 500 PPM is the maximum frequency correction
that ntpd can make. If the local clock is worse than that your options
are to:
a. Live with it.
b. Get it fixed or adjust it (adjtimex???)

If you have a drift file, check the value stored in it. Most machines
will have an absolute value less than 50. If the absolute value is
substantially greater than 100 you may have a problem with the local
clock. Ntpd is supposed to be able to cope but I've not yet seen a
local clock that's off by more than 100 PPM.


Some Linux (and Windows) systems have been found to "lose" clock
interrupts. The usual symptom is that ntpd "steps" the clock quite often.


 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Steve Kost » Wed, 17 Oct 2007 02:24:57


Which means that you changed your kernel.

Your new kernel may require one or more of various command-line options
to allow ntpd to work with it.

Or you may need to re-compile your kernel with a lower HZ value.

Or you may need to recalibrate your tick.

One procedure for recalibrating your tick is documented at
http://www.yqcomputer.com/ #Section_9.1.6.

This topic contains links to other methods of adjusting your tick.

--
Steve Kostecke < XXXX@XXXXX.COM >
NTP Public Services Project - http://www.yqcomputer.com/
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Marc Muehl » Wed, 17 Oct 2007 03:38:44

Richard B. Gilbert schrieb:

The PPM of current system accoring /var/log/messages: 5.706
The old logfile showed: 161.758




The drift file of the new system consists: 6.194
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Martin Bur » Wed, 17 Oct 2007 16:56:14

Marc,

first, I've also just upgraded a x86_64 machine to openSUSE 10.3, and see no
problems using the ntpd version with the distribution.




This seems to indicate that the new kernel implements the system clock based
on a different timer API than the old kernel. That new timer API may not
support the particular chipset on your mainboard correctly.

As a test, you might stop ntpd, then set the system time manually running

date -u; ntpdate ptbtime1.ptb.de

Then let the system time freewheeling, and run

date -u; ntpdate -q ptbtime1.ptb.de

in some intervals. The latter command use "-q" to query only. It does not
set your system time, but reports the difference between the remote NTP
server and your system clock. Please report whether the difference is
growing slow or fast.

BTW, while using the pool servers as you have done is generally the
preferable way to run ntpd, for this particular test you should use a
dedicated server like one of the PTB servers in order to yield more
comparable results.


Martin
--
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Marc Muehl » Wed, 17 Oct 2007 18:33:30

Martin Burnicki schrieb:

# date -u; ntpdate ptbtime1.ptb.de
Di 16. Okt 09:08:09 UTC 2007
16 Oct 11:08:13 ntpdate[2586]: step time server 192.53.103.108 offset 3.867615 sec




# date -u; ntpdate -q ptbtime1.ptb.de
Di 16. Okt 09:17:12 UTC 2007
server 192.53.103.108, stratum 1, offset 2.711909, delay 0.06961
16 Oct 11:17:12 ntpdate[2927]: step time server 192.53.103.108 offset 2.711909 sec

# date -u; ntpdate -q ptbtime1.ptb.de
Di 16. Okt 09:33:04 UTC 2007
server 192.53.103.108, stratum 1, offset 7.275004, delay 0.06960
16 Oct 11:33:04 ntpdate[3562]: step time server 192.53.103.108 offset 7.275004 sec



> Please report whether the difference is growing slow or fast.

As you can see, it it growing fast (more than 7 sec in 20 minutes)
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Marc Muehl » Wed, 17 Oct 2007 18:46:16

Martin Burnicki schrieb:

Would it be a solution to run the ntp server on a different machine with
different chipset and let the current one just be a ntp client. And the ntp
drift technique could get to come to grips with it and it will run fine in future?
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Martin Bur » Wed, 17 Oct 2007 19:28:37


I'm afraid that would not solve the problem. Every NTP node is primarily a
NTP client which gets its time from an upstream time source, which can be
another NTP server, a radio clock or GPS receiver, or whatever.

Using that time source it disciplines its own system time, which can then be
made available to other NTP nodes on the network.

Martin
--
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by Martin Bur » Wed, 17 Oct 2007 19:36:55


Oops, that's too much. Maybe the adjtimex tool can help to decrease the
initial drift to an acceptable value.

Unfortunately I'm not familiar with adjtimex since I haven't had to use it,
yet. Maybe you should search the archive of this group (just google ...),
or someone else might jump in here ...

Martin
--
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany
 
 
 

Problem with xntp after reinstall on openSUSE 10.3

Post by beremou » Sat, 01 Dec 2007 18:42:28

got the same thing in SLES9 SP3 !