Never had to really play with this one, but we have 5 webservers which really use time to display products. Well there was a major power-outage and things didn't start till an hour or so later.

Anway, we have a primary and secondary time server (time.domain.com). The way it 'was' was all the webservers had ntpd running, the status showed the drift but never corrected. I tried an ntpdate time.domain.com and error due to sock in use. I stopped ntpd and then it worked.

So the question, is that how it is supposed to work? Do you stick the ntpdate in a crontab to keep up to date? It looks like the ntpd is the server which the clients shouldn't run. The other question is how to do this for the hardware clock also.

Thanks.

The ntp servers run all the time and they do sync up, just not necessarily instantly. Running ntpdate (or manually setting your system time) is a good way to bring things close right away if you know there's a significant discrepancy. Once you've done that, as long as you have your servers configured in your ntp.conf, then just let the server run.

- Arch

I think this might clarify what you're seeing:

http://www.ntp.org/ntpfaq/NTP-s-algo...ASIC-STEP-SLEW

- Arch