Chapter 9 Troubleshooting
© 2017 Harmonic Inc. All rights reserved.
284
Harmonic MediaGrid Release 4.1
Network Time Protocol (NTP)
From 'mdscore' log file:
2000-12-12 00:00:07.918|D|outoktDropOut|[18..41cf] 1389332264079,
2000.12.12 00:00:07.918779
2000-12-12 00:00:07.918|D|outoktNotifyLiaisonOfExclusion|[18..41cf]
notifying liaison of exclusion
0000-12-12 00:00:07.918|D|outoktFree|(18) drop out
2000-12-12 00:00:07.919|D|outoktFree|(18) closing receive socket 22
2000-12-12 00:00:07.919|D|outoktFree|(18) closing send socket 31
2000-12-12 00:00:10.932|E|startFsCore|Unable to start core: MDS not time synch'ed ****
2000-12-12 00:00:10.932|E|main|Unable to initialize fs core: MDS not time synch'ed ****
Network Time Protocol (NTP)
Network Time Protocol, or NTP, is a way of updating the time on a computer to a known time
reference. For the Harmonic MediaGrid this is essential because while all the computers in the
cluster function as a system, they are still autonomous. To synchronize time between
ContentDirectors, the ntp.conf file on all ContentDirectors in the Harmonic MediaGrid are
configured to point to the same NTP reference, which is typically provided by the SystemManager
but could also be a separate NTP server.
The NTP services that run on the ContentDirector are powerful and many utilities are provided for
troubleshooting problems. A few useful commands and options can help resolve some NTP
problems. One is the ntpdate command.
The two most important subcommands to the ntpdate command are the -q (for query) and the -u
(for update). A time source can be queried to determine what the delta time is between the
source being queried and current time on the computer where the command is being run.
ntpdate query example:
[root@CDL1 ~]# ntpdate -q 10.35.99.210
server 10.35.99.210, stratum 2, offset -110.121850, delay 0.02573
20 Jan 11:27:16 ntpdate[32306]: step time server 10.35.99.210 offset
-110.121850 sec
Synchronizing times on ContentDirectors is straight forward operation using a simple command.
However, when the “omcld” service is restarted, if the ntp.conf file in the /etc directory is not
properly configured for all ContentDirectors in the cluster, it is likely that one or more of the
ContentDirectors will not start properly. The following examples show how to manually update
time on a ContentDirector by synchronizing the time from one ContentDirector to another.
ntpdate update example:
[root@CDL1 ~]# ntpdate -u 10.35.134.201
20 Jan 10:19:37 ntpdate[29958]: step time server 10.35.99.100 offset
224239754.222613 sec
Note that because the “ntpdate -u XXX” command forces a jump in the system clock, it could also
potentially cause problems for DHCPD and authentication. The primary way to avoid such
problems is to have NTP configured properly, and have stable NTP servers that keep accurate
time. In addition, when running the ntpdate command, ideally the affected services should not be
running on the ContentDirector. This should minimize potential clock jumps if the ntpdate
command is run against an active Content Director. The following procedure shows the
recommended order of events:
1. # service omcld stop
2. # service dhcpd stop
