I've been tasked to implement ntp on my SCO Unix servers. I have over 600 servers spread across the US in different time zones. Each remote server has network connectivity to a main server here, through their local ISP. (That's how we currently deliver mail to them).
My question is, how can... (5 Replies)
All
What is the best way to keep the system clock synchronized? I have looked at ntp and netdate. Is one good over the other? Basically I want to know if what is the most secure way to keep the system clock insync. netdate will require me to open up some port 37... is this safe? ntp also... (1 Reply)
i have an HP UNIX box w/c acts as ntp server... I tried to change the time plus 8 minutes... the problem is that the other HP UNIX ntp client did not follow the time... when I tried to restart ntp client... using stop start it only sync to the server once... when I issue the command "ntpq -p", w/c... (2 Replies)
We have configured most of 1200 servers with an NTP server.
For the rest of the servers, I have a strange requirement.
I want to synchronize them with NTP time with a lag of 15 minutes. Is it possible ? If yes How ? (0 Replies)
All here, thank you for listening. Now I've set up a Linux NTP server by adding a external windows NTP server in /etc/ntp.conf. Then I start the ntpd daemon. But how often does the Linux NTP server update its time with the external NTP server? I've looked up everywhere but found no information... (1 Reply)
Good morning,
My first post and first visit, so hello. I have been asked to see if one of our Linux boxes can have it's time sync automatically because the person who built the system has told our staff it cannot be done. To me that raised some flags because although I am no Linux expert, I just... (12 Replies)
There is one server in my company where the time is out of sync.
When I checked (compared with other servers whose time is correct), the XNTPD variable in /etc/rc.config.d/netdaemons was set as 0.
I changed this to 1, and tried restarting the xntpd process with :
# /sbin/init.d/xntpd start
... (4 Replies)
Hello,
one of my clients has the following task for me:
To write a Perl script that checks, whether ntp is active on a particular AIX or Linux box.
Obviously the last synchronization is logged in the file ntp.log
So, my task would be find out, when the synchronization SHOULD have taken... (1 Reply)
Discussion started by: Bloehdian1
1 Replies
LEARN ABOUT DEBIAN
radioclkd
RADIOCLKD(1) Network Time Protocol Daemon RADIOCLKD(1)NAME
radioclkd - decode time from radio clock(s) attached to serial port
SYNOPSIS
radioclkd [ -tphv ] device
DESCRIPTION
radioclkd is a simple daemon that decodes the time from a radio clock device attached to the DCD and/or CTS and/or DSR status lines of
serial port of a computer. It is able to decode the DCF77, MSF and WWVB time signals. The received time is then sent to ntpd using the
shared memory reference clock driver. The type of time signal being received is automatically determined. If you have problems getting the
program to work using interrupts, the following command is known to help in many instances. If this fails you can always fall back to the
polling method.
stty crtscts < /dev/ttyS0
Details on a cheap and easy to make device for receiving these time signals can be found at
http://www.buzzard.org.uk/jonathan/radioclock.html
OPTIONS -p, --poll
Poll the serial port for changes of status in the DCD, CTS and DSR lines rather than use interrupts
-t, --test
Enter test mode printing the length of each pulse and the decoded time at the end of each minute on stdout. The time is not sent to
ntpd using the shared memory reference clock driver in this mode.
-h, --help
Print a short synopsis of the command line arguments.
-v, --version
Print the version number and then exit.
CONFIGURATION
Configuration is very simple. Use server 127.127.28.0 in your ntp.conf file for a clock attached to the DCD line, server 127.127.28.1 for a
clock attached to the CTS line, and server 127.127.28.2 for a clock attached to the DSR line. You will also want to use a fudge line on the
server to change the displayed refid.
CALIBRATION
Due to delays in the propogation of the radio signal, it's processing by the receiver board and the latency of the operating system the
time decoded by the receiver will be slightly offset from actual UTC. Typically this delay will be less than 20ms, so unless you are very
fussy about the time, or are using more than one time source, such as a GPS unit, other radio clock or NTP server on the internet you can
ignore this section.
The basics of the calibration procedure is to determine the average offset of the radio receiver, and use the time1 fudge factor in
ntp.conf to bring the receiver as close as possible to the real time. The easiest way of determining the offset of the radio receivers time
is to run it against a reference clock that does not suffer from these problems. The best reference clock would be a GPS unit. This might
be a GPS unit that you don't wish to dedicate to time keeping, or a borrowed unit. If this is not possible you could use a stratum 1 server
on the internet.
The method of calibration is quite simple. We attach the calibration reference clock to the computer and fudge the stratum of our radio
receiver up to say 5. This way we can be sure that ntpd will lock onto the calibration reference clock. We need to make sure that ntpd is
configured to collect peer statistics so make sure we have some lines similar to these in ntp.conf
statsdir /var/log/ntpstats/
statistics loopstats peerstats clockstats
filegen peerstats file peerstats type day enable
After that we restart ntpd and leave it running for several hours. We can then make a copy the peerstats file. The trick is to remove all
the entries before ntpd has come into close aggrement with the calibration reference clock and then run the peer.awk script in the
scripts/stats directory of the ntp distribution. This will give us a mean offset of our radio receivers in milliseconds. This can them be
converted into seconds and added to the fudge line in ntp.conf for our receiver.
The final step is to remove the change in stratum level for our reference clock and restart ntpd. If you move the receiver any significant
distance then you will need to repeat this calibration step. Across the room or around the current building will be fine, but if you move
it to the next town/city then you will need to recalibrate.
IN USE
The version of ntpd that comes with most Linux distributions does not have the shared memory reference clock driver compiled in by default.
This can be identified by checking the logs after ntpd is started. If the shared memory reference clock driver is not compiled in then the
logs will contain warnings about the reference clock driver not being recognized. To compile ntpd with the shared memory reference clock
driver you must specify the --enable-SHM option when running configure.
Neither radioclkd or ntpd ever mark the shared memory segment for deletion. If you stop using the shared memory reference clock driver
therefore any shared memory segments will persist until you reboot or manually delete the segment using ipcrm. The segments can be identi-
fied as the one with key 0x4e545030, 0x4e545031 or 0x4e545032 using the ipcs command.
BUGS
If you are running a kernel with the PPS kit and have a clock attached to the DCD line you may experience lockups. If you encounter this
problem the currently recomended solution is to move the clock to either the CTS or DSR lines.
AUTHOR
This program was written by Jonathan Buzzard <jonathan@buzzard.org.uk> and may be freely distributed under the terms of the GNU General
Public License. There is ABSOLUTELY NO WARRANTY for this program.
Version 1.0 19 Jan 2003 RADIOCLKD(1)