This is more of a note to self, as I assume and hope that you aren’t making the same mistake I did. I have a couple of NTP servers running here, and I obviously monitor them. The only thing I was doing (and I hope that isn’t the only thing you are doing) was checking that the process (ntpd) is running. I was kindly informed by Michael that the clock on my SixXS gateway was somewhere about 120 seconds off. Impossible, I thought: the ntpd on that machine is up! Well, whaddaya know: the clock had drifted, although I don’t know why that happened. Anyhow, I’ve updated by monitoring configuration to check on the actual offset reported by the ntpd on that particular machine. Since I use Nagios/Icinga, the recipe is simple enough with the included check_ntp_time plugin. As is usual in IT, Murphy is right.