John McMonagle wrote:

Bill Davidsen wrote:

  

quoted

John McMonagle wrote:

    

quoted

Have a raid1  backup server that seems to get corrupted.
This is the 3rd time in about a year.
Have 2 other backup servers that were cloned from this one that have
no problems.

Done a couple kernel upgrades recently.
Now has 2.6.18-2 kernel.
It's based on Debian sarge.

It's a low end Intel server motherboard using ata_piix sata driver.
Have another mother board just like doing raid1 with sata drives that
has had no problems but it has a much lighter disk load.
smartctl has never shown any problems.
In /sys/block/md2 did
echo check > syncaction
No error messages but mismatch_cnt  is 1152.
rc0/errors and rc1/errors are both 0.

I'm guessing a hardware problem.
Any suggestions? 
      

Since memory is the easiest to test, I'd try memtest86+ for at least
12 hr. If this were PATA I'd suggest replugging the cables, but it's
lower probability with SATA. Still, probably worth trying.

    

Ran  Memtest86+ for over 18 hours with no errors.
Also have ecc ram.
I can look at the cables next time I'm  there.
Doesn't  sata do some sort of error checking over the cables?
Anything else to try?

I wish I had another reasonable idea, but memory problems are the big 
target. Is the firmware current level (if you can compare against the 
servers which are working well that would really be good). Other than 
that power supply is the only thing even reasonably likely, particularly 
if it happens under heavy load. Have you looked at the BIOS voltage and 
temp reports just to see if they suggest anything?

-- 
bill davidsen [off-list ref]
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979