Re: [Memtest86] error found
From: Steve Wolfe (unt_at_codon.com)
Date: 09/25/04
- Previous message: John-Paul Stewart: "Re: Expanding/rebuilding a RAID0 array"
- In reply to: Chiefy: "[Memtest86] error found"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Sat, 25 Sep 2004 11:27:25 -0600
> Athlon 2500 + single stick 256MB DDR(PC2700) at 166MHz
>
> Had the opportunity to memtest a box that usually runs continuously.
> As one error was thrown up, I memtested the thing again for 24 hours
> which yielded another single error.
>
> As both errors were at different points, do I need to worry?. The errors
> were a genuine surprise as no problems have been encountered in over 4000
> hours of service. The usual swap use is around 50-100MB.
Do you need to worry? I don't know, do you mind if your computer crashes
or has other problems?
Memtest is designed to create situations inside of the memory that can
bring out errors that you may never see. However, someday, they *might*
crop up. And a single memory error/bit-flip can have a wide range of
effects. If in an unused portion of memory, it will do nothing. If it's in
a segment of unused data, nothing. If that bit happens to help store
someone's name, you may have "Bhiefy" instead of "Chiefy". If that
particular bit happens to be storing a pointer, then the effects are
anything from negligible to a page fault. If that bit is storing code for
one of your device drivers... well, you get the picture.
As for the 4,000 hours of service, there have, of course, been memory
errors during that time. It's inevitable, they happen. And it's like
Russian Roulette, not every "pull" is going to cause anything significant.
I've had a pool of redundant servers that have run for over 13,000 hours at
a time without major incident without using ECC memory, but then again,
sometimes probability catches up with me and things happen. Most of the
machines now have ECC memory just because they've been swapped out with dual
Xeon chipsets that require ECC.
steve
- Previous message: John-Paul Stewart: "Re: Expanding/rebuilding a RAID0 array"
- In reply to: Chiefy: "[Memtest86] error found"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|