[mythtvnz] Bit rot?

[Michael Dilger]

AlanP wrote:
> Every three or four months, my mythbox goes belly up. It works if I reset 
> it -- for about twenty minutes, then it locks up. Hard. Reinstalling it and 
> rescueing my settings from the backup (go go Knoppmyth!) usually seems to 
> fix it.
>   
If reinstalling fixes it, then it is probably file system damage.   If 
you get hard hangs and have to press the power button, that does file 
system damage each time.

You could verify this by running fsck on your root partition by booting 
of CD first.

fsck can fix the structure of the file system, but damaged files 
typically remain damaged and "bit rot" continues.  There are ways to use 
backups and rsync or cpio to determine what has "rotted" and spot-fix 
those changes.

To see what hardware is causing the crashes, leave your system on the 
console (if in X, press ctrl-alt-F1) and turn off screen blanking on 
that terminal (setterm -blank 0 -powersave off), then leave it alone 
until it hangs.   If you have it compiling something big and nasty on 
another terminal, that could help set off a freeze event.   It should 
print out a nice machine check exception error message something like this:

    CPU0: Machine Check Exception 000000000000004
    Kernel Panic - not syncing: CPU Context Corrupt

Then you can run that message through a program called mcelog (you may 
have to install that) to get more information on what the message 
means... it may for instance tell you which subcomponent of your 
computer failed.... most likely some piece of the motherboard you cannot 
replace, or memory... but interesting none-the-less.

Practically, the solution is what others have mentioned:  blow off the 
dust with an air compressor,  upgrade your fans (if you use quiet fans, 
sorry, get loud fans),  down-tweak your BIOS settings (underclock 
things, increase memory timing), and pray that this will be enough.   If 
not, start overvolting as well ... overvolting wears things down faster, 
and creates more heat, but does bring stability.

If you still get hard crashes, you will still get "bit rot" and the only 
solution then is to get a new computer.

Mike