[NTLUG:Discuss] SuSE 9.2 Updates -- LOOKOUT!

[Kevin Brannen]

Stephen Davidson wrote:

> ...
> Hi Chris.
>
> With the Driver, unfortunately no I can't be more specific.  And the 
> corrupted driver is just a guess, but based on some observation and 
> experimentation.  Basically, what is happening is at fixed intervals 
> (I forget if it is every 3 sec's or every 5 sec's), the system "locks 
> up" for 1/2 sec.  During this time, CPU load for the system spikes to 
> %100.  And anytime I try to do anything that requires any kind of work 
> on the system, it is slow and sluggish.  Just bringing up Eclipse (an 
> IDE) runs the system load up to 5 - 7.  A project build can run the 
> load as high as 25 and take 1.5hrs or more.  (Build of same project, 
> in 9.1 before the upgrade, system load of 4, time elapsed, < 5 
> minutes, and that was BEFORE a mess of code in the project was 
> refactored out into another project!).
>
> What is really bizarre is that fact that when the system is cold 
> booted, it behaves normally until the first time a load is put on it.  
> Short loads are no problem.  Longer loads trigger the above 
> behaviour.  I don't have what the "cutoff" is, but normally, just 
> logging in to KDE does not trigger this (unless something causes KDE 
> to take longer than normal, such as another "heavy" process running - 
> in which case starting KDE will take 3 - 5 minutes to start).

...

In a console/xterm, run "top" at the beginning before you stress the 
machine.  Now start stressing the machine and watch the top output.  
Whoever is causing the load should jump to the top.  That info *should* 
be enough to know where to go from there.

HTH,
Kevin