[Gllug] Diagnosing hardware faults

general_email at technicalbloke.com general_email at technicalbloke.com
Fri Dec 3 01:34:14 UTC 2010

On 29/11/10 22:08, John Edwards wrote:
> On Mon, Nov 29, 2010 at 09:48:30PM +0000, Steve Parker wrote:
> <snip>
>> Thanks John E. I was asking more about the comment about memtest; We use
>> mcelog as suggested by John H, and it does seem to pick up a fairly high
>> number of errors (currently got 5 machines out of ~120 reporting
>> errors), but I have recently introduced memtest too, in the belief that
>> it would be a more accurate and thorough test; John E's comment
>> suggested that it might not be all that useful at detecting errors...
> Sorry if I was not clear.
> Memtest very useful in finding errors within the RAM stick. But
> it does not test the RAM when the rest of the server components
> are under load and so may not find some obscure problems.
> Some problems do not occur when each element of a system is tested
> in isolation, only when they are combined or placed in certain
> enviromental conditions. Drives engineers crazy.
> ;)
> Back to the Shuttle boxes. They are useful bits of kit, especially
> for crowded workplaces, but they seem to have a lifetime of about
> 4 years.

Seconded, I've probably seen as many dead ones as I have live ones over 
the years!

On memory testing I wouldn't bother with software, swap out the chip, 
run it as you normally would and see if the problem goes away.

With those symptoms it might well be your PSU though, London power seems 
to wear PSUs down pretty fast and in small systems like the shuttle they 
are pretty weedy little things to start with. Hook it up to something 
with a lot of headroom and see if it fares any better. I've personally 
switched to buying no-brand models rated several hundreds of watts more 
than I technically need and haven't had a single problem since (touch 
wood!!!) but I appreciate a shuttle case might limit your replacement 
options if it is the PSU that's flaky.

It might also be worth check for bulging/leaking capacitors too if 
you've had it a long time and of course giving your disks a scan with 
seatools for dos is rarely a bad idea.


Gllug mailing list  -  Gllug at gllug.org.uk

More information about the GLLUG mailing list