ntp gone wild
Posted: Mon Dec 21, 2009 5:04 am
i have two fit-PC2s side-by-side. i updated both from 8.04 to 9.10. one works fine; the other keeps having system clock crazy resets (e.g. jump forward 9min 59sec) which plays hell with my timer-driven applications
on both, ntpd points to time.ubuntu.com. on the one that goes wild, i have several times done ntpupdate manually, several times done hwclock --systohc, no improvement: for an hour or so, all looks fine; then ntpd jumps the clock forward and system goes into death spiral - timers stop, wlan0 stops (eth0 keeps going), ...
any suggestions on a fix?
the only big anomoly i see in /var/log/syslog is " kernel: [16.856093] mtrr: base (0x3f800000) is not aligned on a size(0x10000000) boundary" which makes me wonder if there is a mmap overlap or something - except i see same entry in the fit-PC2 that works fine
on both, ntpd points to time.ubuntu.com. on the one that goes wild, i have several times done ntpupdate manually, several times done hwclock --systohc, no improvement: for an hour or so, all looks fine; then ntpd jumps the clock forward and system goes into death spiral - timers stop, wlan0 stops (eth0 keeps going), ...
any suggestions on a fix?
the only big anomoly i see in /var/log/syslog is " kernel: [16.856093] mtrr: base (0x3f800000) is not aligned on a size(0x10000000) boundary" which makes me wonder if there is a mmap overlap or something - except i see same entry in the fit-PC2 that works fine