NTP Restarts Cyclically

[Jack Trachtman]

OpenVMS V8.3
TCPIP V5.6 - ECO 5

On most (but not all!) hosts, NTP restarts every five days! I can't find any information in any logs except the OPERATOR.LOG file.

Can anyone tell me what would cause this?

Example OPERATOR.LOG excerpt below:

%%%%%%%%%%% OPCOM 15-DEC-2010 14:41:03.26 %%%%%%%%%%%
***************
Message from user INTERnet on xxx
INTERnet ACP NOLISTEN Process creation success: Service - NTP

%%%%%%%%%%% OPCOM 20-DEC-2010 14:50:28.18 %%%%%%%%%%%
***************
Message from user INTERnet on xxx
INTERnet ACP NOLISTEN Process creation success: Service - NTP

%%%%%%%%%%% OPCOM 25-DEC-2010 14:58:27.36 %%%%%%%%%%%
***************
Message from user INTERnet on xxx
INTERnet ACP NOLISTEN Process creation success: Service - NTP

%%%%%%%%%%% OPCOM 30-DEC-2010 15:01:27.23 %%%%%%%%%%%
***************
Message from user INTERnet on xxx
INTERnet ACP NOLISTEN Process creation success: Service - NTP

%%%%%%%%%%% OPCOM 4-JAN-2011 15:01:37.35 %%%%%%%%%%%
***************
Message from user INTERnet on xxx
INTERnet ACP NOLISTEN Process creation success: Service - NTP

[John Gillings · Accepted solution]

Jack,

This looks normal to me.

NTP log files are (by default) in SYS$SPECIFIC:[TCPIP$NTP]TCPIP$NTP_RUN.LOG

I looked at one system OpenVMS Alpha V8.3 TCPIP V5.6 - ECO 5. There were no OPCOM messages (not enabled), but the NTP_RUN.LOG files were rolling over every day at roughly the same time.

I had a quick look at TCPIP$NTP.CONF (text file in same directory, or see logical name TCPIP$NTP_CONF). There didn't seem to be any time settings.

Is there a problem with the process restarting itself every so often? Remember it comes from a world in which process creations are very cheap, and the general philosophy is to spin off new processes to do just about anything. There's no particular virtue in keeping processes alive forever. Restarting is a simple way to roll over log files, limiting disk consumption.

If you're concerned, check ACCOUNTING for completion status of the processes. Deal with any failures accordingly. Otherwise I suspect that's just the way it works. You could disable the process creation messages if they bother you. My guess is the restart may be based on log file size.

[Jack Trachtman]

John,

I also see that the TCPIP$NTP_RUN.LOG files get created every day. I don't know of any other process on VMS that gets restarted cyclically like this - and w/o any text in the log files stating that this is normal activity or anything in the .CONF file.

I will open a case w/HP to get clarification.

Thanks

[Edwin Gersbach_2]

Jack

I also see a new NTP logfile every day, but the processes TCPIP$NTP_1 on our systems are as old as half a year and more.

You may have some kind of scheduler which restarts NTP every 5 days.

Edwin

[B Novak]

I see the same thing that John sees. Our log file turns over roughly every 24 hours within a +2 to +3 minute window.


TCPIP$NTP_RUN.LOG;2699
10-JAN-2011 22:47:22.21
TCPIP$NTP_RUN.LOG;2698
9-JAN-2011 22:44:45.42
TCPIP$NTP_RUN.LOG;2697
8-JAN-2011 22:42:08.30
TCPIP$NTP_RUN.LOG;2696
7-JAN-2011 22:39:32.51
TCPIP$NTP_RUN.LOG;2695
6-JAN-2011 22:36:50.66
TCPIP$NTP_RUN.LOG;2694
5-JAN-2011 22:34:14.90

[Craig A Berry]

Any chance you're getting the "kiss-of-death" packet where the server essentially requests the client to stop communicating and reset itself?

http://h71000.www7.hp.com/doc/83final/6526/6526pro_031.html#heading_13.4.2.2.1

Have you tried higher levels of logging?

http://h71000.www7.hp.com/doc/83final/6526/6526pro_031.html#ntp_log_sec