I can start ntpd (with /etc/init.d/ntp-server start). ntpq -p shows that it is running (and I can see ntpd in ps output). But after a while (a few minutes), when I run ntpq -p I get the response "Connection refused" -- ntpd is no longer running. It is supposed to run continuously so that it keeps the clock synchronized. I have 3 server entries:

server dewey.lib.ci.phoenix.az.us
server decimal.lib.ci.phoenix.az.us
server pool.ntp.org

When I run ntpq -p now, I get

remote refid st t when poll reach delay offset jitter
==================================================
1.0.0.0 .INIT. 16 u - 1024 0 0.000 0.000 4000.00

Looks as if ntpd is stuck in an initialization operation. That doesn't always happen. I have seen my servers or sometimes some other server (presumably from the pool) specified in the first column.

Any idea why ntpd quits?

I suspect because there is an error in your configuration file, or the ntpd user can't write the lock/drift files because of permission problems

It appears that the problem has to do with time zones. I found a message in /var/log/ntpd that says, "time correction of 21597 seconds exceeds sanity limit." That time correction is about 6 hours, the difference between the time here (MDT) and UTC. I don't think I did anything, but I noticed a short time later that ntpq -p was reporting some success in synchronizing. However, my system clock is now set to UTC. If I reset the time zone using the KDE Control Center, it simply reverts to UTC. ntp seems to be working fine now -- I am also getting message in /var/log/ntpd about the time being reset by a few seconds -- but I am in the wrong time zone.

I seem to have corrected the timezone problem with

cd /etc
ln -s /usr/share/zoneinfo/America/Denver localtime