The Planet Forums > But S.M.A.R.T. works ?

The Planet Forums > System Administration > Server Hardware

giorgiod

Mar 24 2007, 01:25 AM

I all,
recently I had a problem with a SCSI HD on my new dual Opteron box, one of the two disks showed bad signs confirmed by a badblock check.

A disk failure is "normal" but why smartd doesn't send me any warning e-mail ?

These are disk symptoms:

CODE

scsi0: ERROR on channel 0, id 0, lun 0, CDB: Read (10) 00 01 b0 c3 8e 00 01 00 00
Info fld=0x1b0c451, Current sda: sense key Medium Error
Additional sense: Unrecovered read error
end_request: I/O error, dev sda, sector 28361728
scsi0: ERROR on channel 0, id 0, lun 0, CDB: Read (10) 00 01 b0 c4 4e 00 00 08 00
Info fld=0x1b0c451, Current sda: sense key Medium Error
Additional sense: Unrecovered read error
end_request: I/O error, dev sda, sector 28361808

This is the result of the disk check:

CODE

badblocks -v -v -v /dev/sda
Checking blocks 0 to 71687372
Checking for bad blocks (read-only test): 141808644/ 71687372
141809044/ 71687372
141809055/ 71687372
141809066/ 71687372
141809077/ 71687372
141809088/ 71687372
141809099/ 71687372
141809100/ 71687372
141809111/ 71687372
done
Pass completed, 9 bad blocks found.

And this the smartd configuration:

/dev/sda -H -m root -M test -M daily
/dev/sdb -H -m root -M test -M daily

So please, what is wrong in this config ?

Thanks in advance

eth00

Mar 24 2007, 08:36 AM

Well first off does smartctl -a /dev/sda even show an error - it may not. Smartctl is nice but its not all that accurate. I have seen plenty of disks fail without a warning from smartctl, I also have an old celeron at ev1 that complains all the time via smartctl and is still going strong.

I would be more worried the badblocks failed and get the drive swapped out soon. The only possible saving grace is if the badblocks seem to be switched every time it may just be the cable - if they are the same every time it is most likely the drive. I am not sure on the exact procedures on new servers but I don't think hardware tests are always run on them as "new" things do have problems from time to time.

This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.