TITLE: Using chrony to maintain system time
LFS VERSION: 3.1 (should apply to all)
AUTHOR: David Ayers <fogey@mindspring.com>
SYNOPSIS:
Chrony is a software package for maintaining the accuracy of computer
system clocks. It consists of a pair of programs: chronyd is a daemon
running in the background and chronyc is a command line control and
monitoring program.
This hint describes the use of chrony to keep the system clock on a
single computer in step with NTP time over an Internet dial-up
connection. The chrony.txt file referenced below also describes how to
use chrony on systems with a full-time Internet connection or even no
connection at all.
HINT:
Changelog
---------
Information sources
-------------------
chrony.txt located in the chrony-1.17 source tree provides very
complete documentation on the installation and use of chrony.
The man page for 'hwclock' tells how to query and set the hardware
clock (RTC).
My LFS system
-------------
LFS Version CVS-20020131
Linux 2.4.17
The chrony source
-----------------
ftp://chrony.sunsite.dk/projects/chrony/chrony-1.18.tar.gz
Building and installing chrony
------------------------------
Unpack the source code tarball, cd into the source directory and type:
$ ./configure --prefix=/usr
$ make
$ make docs
$ su
# make install
# make install-docs
Choose your NTP server
----------------------
The names and IP addresses for public NTP time servers, along with the
"Rules for Engagement" for their use, can be found at:
http://www.eecis.udel.edu/~mills/ntp/servers.htm
Unless you are configuring a computer for use as a time server for a
large network, choose your NTP servers from the list of secondary
(Stratum 2) servers. Because I live in Illinois, I chose two servers
located at the University of Illinois in Champaign, IL. You will need
both the name and IP of the server(s) you select.
Configuring chrony
------------------
Create the configuration file /etc/chrony.conf.
* # /etc/chrony.conf for LFS
* # Server names and IPs in /etc/hosts.
server [hostname] offline
server [hostname] offline
logdir /var/log/chrony
driftfile /etc/chrony.drift
keyfile /etc/chrony.keys
commandkey 1
dumponexit
dumpdir /var/log/chrony
rtcfile /etc/chrony.rtc
Create a one-line keyfile /etc/chrony.keys. Pick an arbitrary
keyword. (The combination of a numeral and a keyword allows systems to
offer varying levels of access to users as explained in chrony.txt.)
1 [keyword]
Create the log directory and rtc file.
# mkdir /var/log/chrony/
# touch /etc/chrony.rtc
/etc/hosts
----------
Add the IP address and host name of the NTP server or servers selected
to your /etc/hosts file. This is necessary for dial-up connections,
because DNS isn't available to resolve the server names until after
the online connection is established. Name resolution is provided
by the hosts file.
Starting chronyd
----------------
The start command for the chrony daemon is:
# /usr/bin/chronyd -r -s
The '-r' causes chronyd to reload the sample histories stored by the
'dumponexit' and 'dump' commands.
'-s' causes chronyd to set the system clock from the RTC. When used
with '-r' chronyd will attempt to preserve the old samples after
setting the system clock from the RTC. This allows chronyd to perform
long term averaging of the gain or loss rate across system reboots
and is useful for dial-up systems that are shut down when not in use
or when dual booting several OSs.
To start chrony on boot, create an appropriate boot script. I use the
'simple-init' boot system. My boot script:
#!/bin/sh
#
# start chrony service
#
#locate and source functions script
source `PATH=$INIT_D/:/sbin/init.d:/etc/init.d:/etc/rc.d \
type -p functions`
case "$1" in
start)
need $mounted_filesystems || exit_if_any_error
echo 'Starting chronyd daemon...'
loadproc chronyd -r -s || exit_if_any_error
;;
stop)
echo 'Stopping chronyd daemon...'
killproc chronyd
;;
*)
exit 1
;;
esac
exit 0
Using the chronyc control and monitoring program
------------------------------------------------
Type 'chronyc <command>' to monitor the operation of chronyd. The
monitoring commands are:
rtcdata
sources
sourcestats
tracking
For help in interpreting the output of these commands, see chrony.txt.
Internet dial-up configuration
------------------------------
The "offline" flag in chrony.conf prevents chronyd from trying to
query the time service when the computer is not connected to the
Internet. To switch to 'online' when connected, add these lines to
/etc/ppp/ip-up or whatever script takes you online:
/usr/bin/chronyc << EOF
password [keyword]
online
EOF
To return to the offline state when disconnecting, add these lines to
/etc/ppp/ip-down (or equivalent script):
/usr/bin/chronyc << EOF
password [keyword]
dump
writertc
offline
EOF
The driftfile
-------------
One of the main activities of the 'chronyd' program is to work out the
rate at which the system clock gains or loses relative to real
time. Every time this rate information is computed, it is recorded in
the driftfile /etc/chrony.drift. Data in the driftfile looks
something like:
$ cat /etc/chrony.drift
14.6433 1.2223
The first number is the rate at which the system clock gains or loses
time in parts-per-million. The second rate is an estimate of the error
bound around the first value in which the true rate actually lies.
Synchronizing the hardware clock (RTC)
--------------------------------------
The computer relies on the real-time clock (RTC) to maintain the
correct time during shutdown and offline periods. At boot up, chronyd
sets the system time to RTC time. When the Internet connection is
made, chronyd starts measuring the error and drift rate between system
and RTC time, gradually bring both times into sync with the NTP
time signal. After a number of hours online, both system time and RTC time
will equal true time.
My LFS system used a 'setclock' boot script to set system time to the
resl-time clock. Since chronyd now fulfills this function, I removed
'setclock' from the boot process. Any program other than chronyd that
adjusts the real-time clock can confuse the time synchronization
process.
---