load testing suggestion

[Edward]

Reuben D. Budiardja wrote:

> Hi all,
> I have a possible hardware problem with causes some stability problem. So far 
> this stability problems (crashes and locks up) are triggered when I ran rsync 
> for backup-ing remote machines on a ~30- 40 GB data.
> 
> I am tring to determine if this problem only due to disk activity, which 
> indicate disk related hardware problem, or something else, like CPU or 
> motherboard. 
> 
> So I am trying to put this server under heavy load, to test it. Anyone have 
> suggestion on how to do that? I've put things like seti @ home to run, but 
> that program only eat a lot of CPU activity, and does not result in very 
> heavy load (when you do "top" or "w"). The machine is a dual processor.
> 
> Second question, where can I learn more about what the "load average" in top 
> or w really mean? (man top, man w doesn't really help me, or am I looking at 
> wrong place) 
> 
> Thanks in advance for any suggestion.
> 
> PS: I've check for memory problem by running memtest86 on all tests, it's now 
> about 80% on the last test which takes very very long, and no erros have 
> showed up yet. So memory is probably not at fault, at least for now.

When you run memtest86, make sure you change the test suite to include 
all tests. This is something that will need to run overnight (takes a 
very, very long time), but I've had it before that the default test 
suite misses thing the full suite of tests does not.

As for putting a real load on your system, I find compiling a kernel 
will certainly do that.

Regards,
Ed.