Problem: my desktop system hangs. Question: what to do next

Lennart Sorensen lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org
Wed Jul 25 17:36:13 UTC 2012


On Wed, Jul 25, 2012 at 12:47:40PM -0400, D. Hugh Redelmeier wrote:
> | From: Mel Wilson <mwilson-Ja3L+HSX0kI at public.gmane.org>
> 
> | I have a desktop system that can't get through a build of gdb-7.4.1
> | without locking up the desktop, and, apparently, all other processing.
> 
> PC systems can go flakey so many ways.  They don't usually tell you
> why.
> 
> Googling is great but your system's symptoms are so non-specific that
> I cannot see it helping until you get something more.
> 
> I've fixed a small number of PCs -- I'm not a pro.
> 
> It is easy and cheap to run memtest86+ overnight and see what it sees.
> Memory is the problem more often than one would think.  If it isn't the
> problem, it is nice to know that too.  Memtest86+ seems to be a boot
> option on my Ubuntu 10.04 system, so you don't have to look far to
> find it.

Would be nice to know, but it can't tell you that.  memtest can only
tell you that your ram is bad.  It can not tell you that it isn't bad
since you might just not be testing in the manner that shows the failure.

> There is a bunch of stuff to do temperature monitoring (lmsensors).
> I've never figured it out and each motherboard has a slightly
> different configuration.  It sounds as if it could be useful for you.

Certainly in really bad cases if the motherboard has voltage monitors,
those can be useful.

> It is *really* handy to have spares to swap in and out to see if it
> makes a difference

Certainly the most effective method in general.

> Power supplies fail for me more often than I'd expect.  Sometimes
> subtly.  A spare power supply is nice to have.  I admit that it takes
> an annoying amount of work to swap one.  So I put it off to well down
> on my list of experiments.
> 
> My memory isn't what it used to be.  So I keep a lab book to record
> details so that I can look back to see any pattern that forms.  Of
> course it is hard to know what details are worth observing and
> recording.

I have had memory sticks fail and I have had power supplies fail
(although in my situation they usually fail completely, but I have seen
cases where they were just not working right).

> Can you do the problematic build, but switch (before the lockup) to
> the old-fashioned text console?  There is a chance that kernel oopses
> or panics might show up there (but be hidden by a normal X server).
> There are also SysReq key combinations that do kernel diagnostic
> things, best done from a text console.  See, for example,
> <http://www.debian-administration.org/article/457/The_magic_sysreq_options_introduced>

Serial consoles are handy so you can capture on another machine anything
the kernel prints.

> Do tell us what you learn.  Experience is the best teacher; other
> people's experience is cheaper than ones own.  Negative results
> (non-results) are useful too, and are very much under-reported.
> 
> A P4 system isn't a great prize.  At some point it is better to cut
> and run than solve a problem.  This should be especially obvious if
> you are actually paying someone for the problem solving.  This is
> advice that I ignore way too often.  Way, way too often.

I am upgrading my mythtv box tonight since the hardware in it has been
crashing way too often.  I suspect ram/cpu/motherboard problem.  So it
is going from a Q6600+Asus P5K(-R)+4GB to a i7-3860+X79 sabertooh+32GB.
I hope it stops crashing (and gets a wee bit faster too).  My desktop
can receive the old parts to replace the athlon 2800.  It is allowed to
be a bit unstable if I can't figure out the cause.

Doing swapping of DDR2 ram, socket 775 cpu or a mainboard isn't really
an option since I don't have any others of any of those.  I have a machine
with DDR-400, one with DDR3, and a couple with laptop ram.  No help there.
Also the crashes can go a few minutes, a few hours, or a few days,
or even sometimes a few weeks or months.  Completely unpredictable.
I do think it happens more when it is hotter, but I haven't kept any
logs so who knows.

These things are complicated to figure out after all.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists





More information about the Legacy mailing list