dah@ZOOT.AVGRP.CR.ROK.COM (Dave Haverkamp) (01/17/91)
Hello,
Recently (today!) I had a problem with the ntpd daemon. We lost our
connection for a period of a couple hours to the internet. The daemon
kept trying to query the host on the internet almost immediately after
each previous poll. The end result was the ntpd daemon used enough CPU
time that the SUN3 was not able to respond to anything else. ( no login's
etc.). I would like to increase the amount of time that the server
delay's between each poll. (A better fix would be to back off the poll
time if a machine is not responding, instead of constantly trying to
re-establish the connection.)
In ntp.h:
#define NTP_MINPOLL 6 /* (64) seconds between messages */
#define NTP_MAXPOLL 10 /* (1024) secs to poll */
Should I increase the value of both? or should I only change the value
of NTP_MAXPOLL? I also have a SUN Sparc1+ that is running ntpd. The
time daemon noticably degraded the preformance on it also. On the
Sparc I was able to login and kill the ntpd daemon though.
----> $Revision: 3.4.1.7 $ $Date: 89/05/18
I do have the xntp files, but have not looked at them to see if this problem
exists with them also.
Thank you for your time.
David Haverkamp
Rockwell International
Collins Commercial Avionics
400 Collins Road NE
Cedar Rapids, Iowa 52498
(319) 395-8475Mills@udel.edu (01/17/91)
Dave, Something is broken. No chimer should ever chime more often than 2**NTP.MINPOLL seconds in any case. If it does, maybe something is broken in the system timekeeping code. Note that very old (Version 0) ntpd had a problem in responding to polls, but not in initiating them. I think all those old versions (circa 1987) are gone. Dave