So, I guess OpenSolaris is a better memtest than memtest86.
On Fri, Jun 5, 2009 at 7:44 AM, Chris Wells <chris.unix.dude at gmail.com>
wrote:
> Thanks to all who responded.
>
> I didn't have a comparison system, but I was able to isolate the
> problem down to a single DDR2 stick (via trial and error).
> annoyingly, even once I'd which stick was to blame, even when I only
> used that stick, memtest86+ did not find any errors on
> it.
>
> Cumulatively, I had left memtest86 running for over 40 hours, and it
> completed 24 full runs over that period. (memtest86+ 2.11).
>
> Still, I don't have way of putting load onto the system to guarantee a
> failure, so sometimes the system would be stable for 24 hours, and on
> other occasions (worst case!) 24 seconds.
>
> The best way of loading up the machine I found was to run "zpool
> scrub" at the same time as running a couple of VMs (in VirtualBox of
> course), with a couple of diagnostic tools running inside.
>
> And it was interesting that Windows7 didn't crash at all (and
> continued running two benchmarks simultaneously over 24 hours). ?I
> guess maybe a lots of Windows 7 is bloat, and the affected physical
> memory location happened to coincide with data, not program stack, or
> that Windows doens't actively check NTFS filesystem checksums like ZFS
> does.
>
> Anyway - problem found.
>
> Chris
>
>>
>> Already mentioned but just to be sure.. a barebone setup (board, CPU, 1
>> stick of RAM, graphics) is best - I don't know if your board has onboard
>> video but you might like to try and revert to that for testing, OTOH,
>> disable all south bridge functions (nic, sound, etc) - There's often a 'load
>> failsafe defaults' option in the BIOS setup.
>>
>> As mentioned, check the cable. Actually don't bother checking it, just
>> replace it.
>>
>> What are your temps like (CPU/case)? Is cooling adequate? Does the ambient
>> environment seem to effect the frequency of crashes? Adequate power supply
>> to the components is also important - are you using a nice weighty high
>> wattage PSU? You mention high IO load raises the issue - maybe there isn't
>> enough juice?
>>
>> Might like to check a non Solaris based test suite as well as the Seagate
>> Seatools suite (will pass though I'm sure), ultimate boot cd (UBCD) has a
>> few other disk testing utilities included.
>>
>> Good luck,
>>
>> Peter Davidoff.
>
>
>
> --
>
> Regards,
>
> Chris
> _______________________________________________
> ug-msosug mailing list
> ug-msosug at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/ug-msosug
>
>
