[lug] Server clock losing serious time -- SOLVED

[Jeff Schroeder]

Heh, nothing like answering your own post...

It turns out the system (software) clock was having serious issues 
because the hard disk wasn't configured to use DMA.  It's the famous 
"losing too many ticks!" error you see at boot time (in my case, via 
'dmesg') where the kernel decides the clock isn't trustworthy.

Although I don't fully understand the kernel-hardware clock interface, 
when I reconfigured the hard disks and rebooted, the clock settled down 
and behaved.  Moreover, NTP appears to work as expected.

Note to self: check ALL system logs, including boot logs, when 
troubleshooting stuff! :)

Jeff