[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#252289: general: system clock goes chaoticly causing system breakdown

On Tue, 08 Jun 2004, Antos Andras wrote:
> In the logfile I see that the ntpdate offset increases only 0.08 or 0.16 
> seconds an hour while I am not logged in, but it can be anything up to 150 
> seconds an hour while I am logged in. (Shell I send the logfile?)

Yes, please.  But apt-get --purge remove the following packages before you
produce that log (and cold-boot the machine after removing the packages):


and rm -f /etc/adjtime

(that should take care of the stuff that might be mucking with your RTC and
system time on the background).

Also, please describe exactly WHAT you do when you are logging in locally.
Stuff of interest:

  Running an X server (what X server, what driver, what DRI driver)

  Anything that uses the RTC (mplayer comes to mind)

  Keep an eye on /proc/interrupts, and make sure the timer interrupt is
  correctly increasing at 100HZ (especially when you notice the timer
  "is stuck".

  Make *extremely* sure to always say
    "system time/clock" for the stuff you get from date and gettimeofday()
    "RTC time" for the stuff you get from hwclock -r
  I for one don't exactly know which of your clocks is doing weird
  things (the system clock, or the RTC).

If for some reason your kernel uses something else than 100 for HZ, you
should get that number of timer interrupts per second.  And in that case,
NEVER install or run adjtimex in that machine.

  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

Reply to: