On 2006-12-17 00:00, ByteEnable wrote: > <snip> > Yeah, I deleted the /etc/adjtime file, played with hwclock, played with > ntp, etc. Its just no workie. <sigh>. > > Byte > > Stop ntpd, delete that file *and* /var/lib/ntp/drift/ntp.drift, then:
grep USER_HZ /usr/src/linux/include/asm/param.h adjtimex -p In the latter, the value of "tick" should be 1 million/USER_HZ. If not, then run: adjtimex -t <val> -f 0 -o 0 with <val> equal to that value. Now running "adjtimex -p" again should show that "tick" has that value, while the values of "offset" and "frequency" are zero. The system clock will now be running in an uncorrected, "raw", state. Use "hwclock --hctosys" to bring the system clock into agreement with the hardware clock. Now run "adjtimex -c" to determine any discrepancy between the rates of the system and hardware clocks. Comparisons are 10 seconds apart, and after the first two, adjtimex will suggest values of "tick" and "frequency" that will keep the two clocks at the same rate. Also use "adjtimex -h <timeserver>" at least twice, over a period of at least a couple of hours (the longer the better). Do not reboot the system, do not run the ntp daemon, and do not set any time variables in either system or hardware clocks, during this time to ensure accurate calculations. If /var/log/clocks.log exists, delete it before the first run. Starting with the second run, adjtimex will calculate errors in the frequency, in ppm, for each of the system and hardware clocks. It's probably a good idea to let us have a look at the results of all this at this point. -- The best way to accelerate a computer running Windows is at 9.81 m/s² -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
