Configure the clock source

OceanBase products have implicit relationships in time. Therefore, the physical time on all servers of the OceanBase products must be synchronous with an offset of less than 100 ms or even smaller. Therefore, the clocks of multiple nodes in a cluster and OCP nodes must be configured with the clock synchronization service (NTP or Chrony) to ensure that the clock offset of all nodes is within 100 ms. This topic describes how to configure the clock source under the Network Time Protocol (NTP) for multiple servers in a cluster. Skip this topic if you install OceanBase Database on a single server.

Note

The clocks of servers in the OceanBase cluster must be synchronized. Otherwise, the OceanBase cluster may not start or will encounter operational exceptions. A physical server is synchronous with the clock source server when the clock offset between them is 50 ms or lower. The tolerable clock offset of an OceanBase cluster cannot exceed 100 ms. When the clock offset exceeds 100 ms, the cluster may become leaderless. You can synchronize the clock and restart the OceanBase cluster to restore it to normal. If you have already configured the NTP clock source, no need to do it again.

Prerequisites

You have the root user privileges on all OBServers.

Procedure

Perform the following steps to configure the NTP clock source.

Note

Generally, the built-in NTP service of Linux can be used for time synchronization. If the current environment has a stable and reliable NTP server, this NTP server is used as the NTP source of all servers. If no such NTP server is available, a fixed server is selected as the NTP source.

1. Install the NTP source on each server through the YUM package manager.

   [root@xxx /]# yum install ntp ntpdate -y
   

2. Configure the ntp.conf file.

   [root@xxx /]# vi /etc/ntp.conf
   restrict default ignore
   restrict 127.0.0.1
   restrict 192.168.0.0 mask 255.255.0.0
   
   driftfile /var/lib/ntp/drift
   pidfile /var/run/ntpd.pid
   #logfile /var/log/ntp.log
   
   # local clock
   server 127.127.1.0
   fudge 127.127.1.0 stratum 10
   
   server 192.168.XX.XX iburst minpoll 4 maxpoll 6
   

   Parameters:

   • restrict specifies the privileges of time synchronization commands related to the IP address.

     restrict [IP] mask [netmask_IP] [parameter]
     

     Valid values of parameter: ignore, nomodify, noquery, notrap, and notrust. ignore specifies to reject all types of NTP synchronization by default.

   • server specifies the upper-level NTP source server.

     server [IP or hostname] [prefer]
     

     if no upper-level NTP source server is available, the server parameter can be set to 127.127.1.0, which means that the local server is used as the NTP source server.

3. Restart the NTP synchronization service.

   [root@xxx /]# systemctl restart ntpd
   [root@xxx /]# systemctl status ntpd
   

4. Check the NTP synchronization status.

   Note

   If the NTP service has just been started, you need to wait for a moment before running the ntpstat command.

   [root@xxx /]# ntpstat
   [root@xxx /]# ntpq
   remote          refid       st  t  when  poll reach delay offset jitter
   ========================================================================
   LOCAL(0)        .LOCL.      10  l  589   64   0     0.000 0.000  0.000
   *192.168.XX.XX  10.10.10.1  2   u  18    64   377   1.591 0.249  0.054
   

   Parameters in the result of the ntpq command:

   • remote: indicates the NTP server used. * indicates the current NTP server. LOCAL indicates that the local server is used as the NTP server. x indicates that the NTP server is no longer used. - indicates that the NTP server is no longer used. + indicates that the NTP server is in good conditions and preferred. # indicates that the NTP server is in good conditions but not used.
   • refid: indicates the higher-level NTP server used by the remote NTP server. INIT indicates that the higher-level NTP server is being obtained.
   • st: indicates the stratum of the remote NTP server.
   • when: indicates the period from the last synchronization to the current time, in seconds by default. h indicates hour and d indicates day.
   • poll: indicates the synchronization frequency, in seconds.
   • delay: indicates the roundabout time from the local server to the remote NTP server, in milliseconds.
   • offset: indicates the time offset between the local server and the remote NTP server, in milliseconds.
   • jitter: indicates the average deviation in the time offset between the local server and the remote NTP server, in milliseconds.

5. Conduct final check.

   Use the following sample code for reference to conduct final synchronization latency check:

   [root@xxx /]# clockdiff 192.168.XX.XX
   ..................................................
   host=192.168.XX.XX rtt=1(0)ms/1ms delta=0ms/0ms Sat Apr 18 14:41:40 2020
   

6. (Optional) Perform manual synchronization.

   If the time deviation is always great, you can use the ntpdate command to manually correct the time. This conflicts with the automatic synchronization service. Therefore, before you manually correct the time, you need to stop the NTP service first.

   [root@xxx /]# systemctl stop ntpd
   [root@xxx /]# ntpdate -u 192.168.XX.XX
   18 Apr 14:54:20 ntpdate[108001]: adjust time server 192.168.XX.XX offset -0.000180 sec
   

   If this method is effective, you can configure it in crontab of the system.

   [root@xxx /]# crontab -e
   * * * * * /sbin/ntpdate -u 192.168.XX.XX 2>&1 1>>/tmp/ntpupdate.log
   

7. (Optional) Configure the firewall rule.

   In a standard environment, the firewall needs to be disabled. If the firewall must be enabled in a server environment, comply with the following rule:

   [root@xxx /]# iptables --A INPUT --p udp --i eth0 --s 192.168.1.0/16 --dport 123 --j ACCEPT
   [root@xxx /]# systemctl restart ntpd