I have been known to lose track of time every once in a while, but not 3 days. I just logged onto our monitoring server after noticing some oddness in time. When logging on, the server said it was the 29th of October. That was 3 days ago. So I execute ntpdate pool.ntp.org and everything was back in sync, but this was odd. I vaguely remembered having to do that shortly after rebuilding the box. So I ran the command again, and sure enough, in a span of about 10 minutes, the server had lost about 92 seconds. Time for some research!
After googling about a bit, I stumbled across this little gem. Apparently due to the nature of how linux keeps the clock in sync, and the fact that the host server is a little busy, this was causing a massive time loss. The article contains details on how to fix it.
After running with the suggested changes for a few hours, I’ve not seen much of a time loss (0.001 of a second) over the few hours. Much better.