个人工具

Quick HOWTO : Ch24 : The NTP Server

来自Ubuntu中文

跳转至: 导航, 搜索


Introduction

The Network Time Protocol (NTP) is a protocol used to help synchronize your Linux system's clock with an accurate time source. There are that allow the general public to synchronize with them. They are divided into two types:


  • Stratum 1: NTP sites using an atomic clock for timing.
  • Stratum 2: NTP sites with slightly less accurate time sources.

It is good practice to have at least one server on your network be the local time server for all your other devices. This makes the correlation of system events on different systems much easier. It also reduces Internet bandwidth usage due to NTP traffic and reduces the need to manage firewall rules for each NTP client on your network. Sometimes, not all your servers will have Internet access; in such cases you'll need a central server that all can access.

For a list of available Stratum 1 and 2 servers consult http://www.ntp.org/


Download and Install The NTP Package

Most RedHat and Fedora Linux software products are available in the RPM format. Downloading and installing RPMs isn't hard. If you need a refresher, Chapter 6, "Installing Linux Software", has all the details.

When searching for the file, remember that the NTP RPM's filename usually starts with the word ntp followed by a version number as in ntp-4.1.2-5.i386.rpm.

The /etc/ntp.conf File

The /etc/ntp.conf file is the main configuration file for Linux NTP in which you place the IP addresses of the stratum 1 and stratum 2 servers you want to use. Here are the steps to create a configuration file using a pair of sample Internet-based NTP servers:

1) First we specify the servers you're interested in:

server  otherntp.server.org    # A stratum 1 server at server.org
server  ntp.research.gov       # A stratum 2 server at research.gov

2) Restrict the type of access you allow these servers. In this example the servers are not allowed to modify the run-time configuration or query your Linux NTP server.

restrict otherntp.server.org   mask 255.255.255.255 nomodify notrap noquery
restrict ntp.research.gov      mask 255.255.255.255 nomodify notrap noquery

The mask 255.255.255.255 statement is really a subnet mask limiting access to the single IP address of the remote NTP servers.

3) If this server is also going to provide time for other computers, such as PCs, other Linux servers and networking devices, then you'll have to define the networks from which this server will accept NTP synchronization requests. You do so with a modified restrict statement removing the noquery keyword to allow the network to query your NTP server. The syntax is:

restrict 192.168.1.0 mask 255.255.255.0 nomodify notrap

In this case the mask statement has been expanded to include all 255 possible IP addresses on the local network.

4) We also want to make sure that localhost (the universal IP address used to refer to a Linux server itself) has full access without any restricting keywords:

restrict 127.0.0.1

5) Save the file and restart NTP for these settings to take effect. You can now configure other Linux hosts on your network to synchronize with this new master NTP server in a similar fashion.

How To Get NTP Started

You have to restart the NTP process every time you make a change to the configuration file for the changes to take effect on the running process.

To get NTP configured to start at boot, use the line:

[[email protected] tmp]# chkconfig ntpd on

To start, stop and restart NTP after booting, follow these examples:

[[email protected] tmp]# service ntpd start
[[email protected] tmp]# service ntpd stop
[[email protected] tmp]# service ntpd restart

Testing And Troubleshooting NTP

After configuring and starting NTP, you should test it to make sure it is working. Here are some guidelines you can follow to get NTP working correctly.


Verifying NTP is Running

To test whether the NTP process is running use the command

[[email protected] tmp]# pgrep ntpd


You should get a response of plain old process ID numbers.


Doing An Initial Synchronization

If the time on the local server is very different from that of its primary time server your NTP daemon will eventually terminate itself leaving an error message in the /var/log/messages file. You should run the ntpdate -u command to force your server to become instantly synchronized with its NTP servers before starting the NTP daemon for the first time. The ntpdate command doesn't run continuously in the background, you will still have to run the ntpd daemon to get continuous NTP updates.

Take a look at some sample output of the ntpdate command in which a server whose initial time was set to midnight, was correctly set to 8:03 am.

  • The date was originally set to midnight which was verified by using the date command.
[[email protected] tmp]# date
Thu Aug 12 00:00:00 PDT 2004
[[email protected] tmp]#
  • The ntpdate command is run three times to synchronize smallfry's clock to server 192.168.1.100, but it must be run while the ntpd process is stopped. So you'll have to stop ntpd, run ntpdate and then start ntpd again.
[[email protected] tmp]# service ntpd stop
[[email protected] tmp]# ntpdate -u 192.168.1.100
Looking for host 192.168.1.100 and service ntp
host found : bigboy.my-site.com
12 Aug 08:03:38 ntpdate[2472]: step time server 192.168.1.100 offset 28993.084943 sec
[[email protected] tmp]# ntpdate -u 192.168.1.100
Looking for host 192.168.1.100 and service ntp
host found : bigboy.my-site.com
12 Aug 08:03:40 ntpdate[2472]: step time server 192.168.1.100 offset 2.467652 sec
[[email protected] tmp]# ntpdate -u 192.168.1.100
Looking for host 192.168.1.100 and service ntp
host found : bigboy.my-site.com
12 Aug 08:03:42 ntpdate[2472]: step time server 192.168.1.100 offset 0.084943 sec
[[email protected] tmp]# service ntpd start
[[email protected] tmp]#
  • The date is now corrected.
[[email protected] tmp]# date
Thu Aug 12 08:03:45 PDT 2004
[[email protected] tmp]#


Determining If NTP Is Synchronized Properly

Use the ntpq command to see the servers with which you are synchronized. It provided you with a list of configured time servers and the delay, offset and jitter that your server is experiencing with them. For correct synchronization, the delay and offset values should be non-zero and the jitter value should be under 100.

[[email protected] tmp]# ntpq -p

Here is some sample output of the command:

     remote          refid       st t when poll reach  delay   offset   jitter
==============================================================================
-jj.cs.umb.edu   gandalf.sigmaso  3 u   95 1024  377  31.681  -18.549    1.572 
milo.mcs.anl.go  ntp0.mcs.anl.go  2 u  818 1024  125  41.993  -15.264    1.392
-mailer1.psc.edu ntp1.usno.navy.  2 u  972 1024  377  38.206   19.589   28.028
-dr-zaius.cs.wis ben.cs.wisc.edu  2 u  502 1024  357  55.098    3.979    0.333
+taylor.cs.wisc. ben.cs.wisc.edu  2 u  454 1024  347  54.127    3.379    0.047
-ntp0.cis.strath harris.cc.strat  3 u  507 1024  377 115.274   -5.025    1.642
*clock.via.net   .GPS.            1 u  426 1024  377 107.424   -3.018    2.534
ntp1.conectiv.c  0.0.0.0         16 u    - 1024    0   0.000    0.000  4000.00


Your Linux NTP clients cannot Synchronize Properly

A telltale sign that you haven't got proper synchronization is when all the remote servers have jitters of 4000 with delay and reach values of 0.

    remote           refid      st t when poll reach   delay   offset  jitter
 =============================================================================
LOCAL(0)        LOCAL(0)        10 l    -   64    7    0.000    0.000   0.008
ntp-cup.externa 0.0.0.0         16 u    -   64    0    0.000    0.000 4000.00
snvl-smtp1.trim 0.0.0.0         16 u    -   64    0    0.000    0.000 4000.00
nist1.aol-ca.tr 0.0.0.0         16 u    -   64    0    0.000    0.000 4000.00

This could be caused by the following:

  • Older versions of the NTP package that don't work correctly if you use the DNS name for the NTP servers. In these cases you will want to use the actual IP addresses instead.
  • A firewall blocking access to your Stratum 1 and 2 NTP servers. This could be located on one of the networks between the NTP server and its time source, or firewall software such as iptables could be running on the server itself.
  • The notrust nomodify notrap keywords are present in the restrict statement for the NTP client. In some versions of the Fedora Core 2's implementation of NTP, clients will not be able to synchronize with a Fedora Core 2 time server unless the notrust nomodify notrap keywords are removed from the NTP client's restrict statement.
In this example the restrict statement has only the client network defined without any keywords and the configuration line that works with other NTP versions has been commented out:
# -- CLIENT NETWORK -------
#restrict 172.16.1.0 mask 255.255.255.0 notrust nomodify notrap
restrict 172.16.1.0 mask 255.255.255.0


Fedora Core 2文件权限许可

所有的Fedora/RedHat NTP 后台进程要写临时文件到 /etc/ntp目录. 不幸的是,Fedora Core 2, 不允许这个写临时文件,所以要设置组和使用者权限给NTP.

[[email protected] tmp]# chown ntp:ntp /etc/ntp

如果你没有这么做,你会在/var/log/messages文件看到下面信息.

Aug 12 00:29:45 smallfry ntpd[2097]: can't open /etc/ntp/drift.TEMP: Permission denied

Configuring Cisco Devices To Use An NTP Server

You can use NTP to synchronize time on a variety of devices including networking equipment. I have included the necessary NTP commands for a variety of Cisco Systems products because it is one of the most popular manufacturers of networking equipment and would feature in the overall architectures of many home office/small office (SOHO) environments and corporate departments.


Cisco IOS

To make your router synchronize with NTP servers with IP addresses 192.168.1.100 and 192.168.1.201, use the commands:

ciscorouter> enable
password: *********
ciscorouter# config t
ciscorouter(config)# ntp update-calendar
ciscorouter(config)# ntp server 192.168.1.100
ciscorouter(config)# ntp server 192.168.1.201
ciscorouter(config)# exit
ciscorouter# wr mem

The ntp server command forms a server association with another system, and ntp update-calendar configures the system to update its hardware clock from the software clock at periodic intervals.


CATOS

To make your router synchronize with NTP servers with IP addresses 192.168.1.100 and 192.168.1.201, use the commands:

ciscoswitch> enable
password: *********
ciscoswitch# set ntp client enable
ciscoswitch# ntp server 192.168.1.100
ciscoswitch# ntp server 192.168.1.201
ciscoswitch# exit

The ntp server command forms a server association with another system, and set ntp client enable activates the NTP client.


NTP Security

You should always be aware of how NTP can be affected by your network's security policy. Here are some common areas of concern.

Firewalls and NTP

NTP servers communicate with one another using UDP with a destination port of 123. Unlike most UDP protocols, the source port isn't a high port (above 1023), but 123 also. You'll have to allow UDP traffic on source/destination port 123 between your server and the Stratum 1/2 server with which you are synchronizing.

A sample Linux iptables firewall script snippet is in Appendix II, "Codes, Scripts, and Configurations".

NTP Authentication

There may be cases where you want to not only restrict NTP synchronization to specific networks but also to require a synchronization password. This is beyond the scope of this book, but is covered in detail at the NTP website www.ntp.org.

Configuring A Windows NTP Client

Windows clients that are part of an Active Directory domain automatically get their time synchronized from the domain server. If your client is not part of a domain you can add your new NTP server to your Windows client. Here's how:

  1. Click on the time at the bottom right hand side of your screen.
  2. Click on the "Internet Time" tab of the dialog box
  3. Click the check box labeled "Automatically synchronize with an Internet time server" and enter the name or IP address in the box underneath it.
  4. Click on the "Update Now" button

You will get a message saying "Your time has been successfully synchronized" when the operation is complete.

Conclusion

It is important that all the systems under your control have the same accurate time. It can help to give a very clear indication of a chain of events that involve multiple devices and it can also help in the synchronization of time sensitive-transactions.

Having an NTP server on your local network can make this easier to do. Sometimes it isn't desirable for all your NTP clients to have access to the Internet to synchronize with stratum 1 and 2 servers, even when they all have access there is the risk of them losing synchronization if the central connection to the Internet is lost. The maintenance of firewall rules for multiple NTP connections to the Internet can also be daunting especially if the management of the firewall is handled by another group.

A local NTP server can ensure that the clients all have the same time relative to the server even when Internet connectivity is temporarily lost thereby reducing the problems of them being out of synchronization with each other. The firewall rules can also be greatly simplified. A local NTP server is frequently a good thing to have for these reasons.