Hi Radu,

thank you for using NTP and our installer! Please see below for a post that 
ended up being longer than intented! Sorry for
this :-) ...

Am 06.07.2011 13:32, schrieb [email protected]:
> Hello,
> 
> I have an issue with the Meinberg software:
> 
> I have defined a simple configuration file 
> 
> Server time1
> Server time2
> 
> I can see be running the ntpq -p command that the client is using the two 
> servers: one as preferred (*) the other one on standby (+) and also the 
> offset which is around 120 seconds. The problem is that Meinberg is not 
> setting up the system clock by itself. However, if I restart the service, 
> it does it at the start, but if I modify the clock afterwards to simulate 
> an offset, it still does not sync system clock.
> 
> Thank you!

Well, you already received a number of good and correct answers so far. I just 
wanted to confirm and summarize a few things,
since you refer to the "Meinberg software" and I feel responsible for it :-) ...

1. Meinberg created the installer for ntpd, which is the product of an open 
source project (see www.ntp.org). As David
Woolley already pointed out, we take the sources from ntp.org, and build a 
binary which we then put into our own installer.
This is an attempt to make it easy for Windows users to install and run ntpd on 
there systems (as Steve pointed out, we
apply a few minor patches to it), and so far we received a lot of positive 
feedback for this. Please note that we still take
responsibility for the whole package and gladly  support to everybody who uses 
it, we just do not want to claim that this is
"our" software in regards to the thousands of hours that volunteers spent on 
this software for tens of years now.

2. It is also true that the test procedure you try to use to evaluate ntpd is 
not covered by ntpd. As a number of people
correctly explained, ntpd has not been designed to correct the system time in 
every situation, especially not when it
detects that someone manipulates it or the time changes are so big that ntpd's 
approach to synchronize the system time are
not working reliably anymore. I have to say that this is a very common 
misunderstanding, and we have a lot of customers and
ntp users who try to test ntpd this way. [note to myself: check support.ntp.org 
if we already have an article which covers
testing ntp and if not, create one] A good reading is the "NTP debugging 
techniques" page of the ntp.org documentation (e.g.
http://doc.ntp.org/4.2.4/debug.html).

3. The two limits affecting your case in terms of offset correction are the 
128ms step threshold and the 1024s panic
threshold. The step threshold basically  means: if the offset is larger than 
128ms, step the time instead of trying to
slowly correct it and "drag" it towards 0. The panic threshold covers very 
large time jumps (often used in test procedures),
it basically says: if the offset grows over 1000 seconds (~ 20 minutes), stop 
everything and abort ntpd. The reason behind
this is that such a large time jump is indicating that someone or something 
manipulates the clock (by accident, deliberately
or due to some hardware/software fault).

The step threshold comes into play in your case (changing the time by 120s), 
meaning that ntpd should step the clock once it
verified that the offset is permanent and not a single spike/mismeasurement. 
This verification takes a while (15 min per
default), after all most people do not want ntpd to quickly kick the system 
time for 1000s when it sees a temporary offset
from its references (i.e. upstream servers or hardware refclocks). This 
verification period is referred to as the stepout
interval.

All three values can be changed using the "tinker" configuration command, but I 
have to stress that their default values
(128ms, 1000s, 900s) have been selected carefully and should not be altered 
just for test reasons. If you come to the
conclusion that in your case you want to change them, please refer to the 
"Miscellaneous Options" page of the ntp.org
documentation, which is installed by our installer, too (if you did not 
de-select it). You can also check the online
documentation at doc.ntp.org, but please make sure that you look into the 
documentation set that corresponds to the version
of ntpd you are using (e.g. for 4.2.4: http://doc.ntp.org/4.2.4/miscopt.html). 
The documentation included in our NTP
installer is of course corresponding to the ntpd version it installs.

Best Regards,
 Heiko

_______________________________________________
questions mailing list
[email protected]
http://lists.ntp.org/listinfo/questions

Reply via email to