Re: many lost ticks, clock drifts problems also with debian smp kernel?
Selon Andrew Sharp <andy@netfall.com>:
> Define extended period of time.
My definition for "extended period of heavy load" would be:
>From 2 days up to 2 weeks of Integer/floating point computation with both cores
maxed out, all 4 Gigs of RAM used up, 10 Gigs of swap allowed, log files of
several Gigs incrementally written on a 500 Go RAID0 array, plus heavy network
load.
Usually the clock drift starts to be really noticeable after a few hours, a week
after the machine still runs and computes the clock is 4 months off, you can't
log into the machine because all passwd timers time out before you can even
move your finger !
By the way the 2.6.14.3 kernel seems to have cured the problem for me, but now
my promise IDE controller and my built in ethernet card don't work anymore...
Reply to: