rudolf wrote:
> I've got a dual-opteron machine here and have some problems with ntpd. 
> Everything is fine when the machine does nothing, but when there is some 
> load (cvs, rsync), it starts to log 'frequency errors'. Is it serious? 
> NetBSD 2.0_STABLE.

It looks like two or more of your upstream NTP sources disagree about time to 
such an extent that they are swinging your clock back and forth.  Either that, 
or maybe your mainboard's hardware clock is busted.

Run ntpq and then the peers command, and see whether two or more servers have a 
significant offset.  I run a ring-of-three NTP servers as peers, each syncing 
to a seperate major timesource (like time.apple.com, nist.gov, etc) and a 
pool.ntp.org random IP:

ntpq> peers
      remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
+thesimonet.org  140.142.16.34    2 u   79  512  377  131.341    1.217   1.188
+NAVOBS1.MIT.EDU .PSC.            1 u  142  512  377   10.561    5.535   0.187
*bonehed.lcs.mit .CDMA.           1 u   75  512  377   10.189    5.419   0.199
+time.apple.com  17.106.100.13    2 u   75  512  377   83.711    4.322   0.496
  shot            199.103.21.227   3 u  291 1024  375    0.367   -3.970   1.191
xpublic          18.145.0.30      2 u  144  512  377    0.441  -13.588   0.091
-ns1             63.73.218.130    2 u  286 1024  176    0.150    1.171   0.291

-- 
-Chuck