I've seen situations where people have put ntpdate in a cronjob to get around issues with big time jumps at boot or dodgy clocks under virtualization. There are much better solutions to this problem, so let us know if this is the case for you.
put "tinker panic 0" in your ntp.conf.
This stops ntpd from throwing a wobbly if it sees large jumps in time and will keep clocks in virtualised environments ticking along.