James Cameron wrote:
Okay, good, that excludes single cause memory or CPU
problems. In my
day job I've seen systems that will not trigger memtest failures yet
show evidence of them when other devices are used. memtest doesn't
exercise the memory at the same time as the PCI bus or devices on it.
So it helps here by suggesting the memory and CPU are fine.
One thing I haven't done, though, is run it for a very long period of
time. I had let it go for... about four hours or so? Somewhere in there.
I'll do that again.
Thanks. Also, take note of what hangs ... does the
mouse move on
screen, does the capslock LED respond on the keyboard, are there any
unusual LED indications from the front panel; e.g. disk LED jammed on.
This tends to vary. Sometimes, the mouse pointer still moves but I can't
click anything. Sometimes, it's a total hard lock, and the lights on the
keyboard for num lock and caps lock won't work at all. Sometimes, it's
just Firefox that locks up, and I have to kill it.
The mouse pointer still moving, though, has been common enough hang that
it stands out in my memory.
There are some really interesting methods to capture
data about a hang,
but not sure if you want to go there yet. My favourite is a serial
console with a laptop next to the failing system.
That sounds like fun. I haven't broken out my laptop (really, an MSI
Wind netbook, but hey, it works and it does what I need it to) since the
last semester ended, so it needs a bit of exercise.
-- Darren