Badblocks and hardware RAID.
 

Hi.

I've had EIDE disks attached in RAID 1 to a 3Ware 7000-2 pci board for several years. Initially, I had 2x80GB disks. At some point the RAID boards started to alarm for bad sectors. I eventually changed the disks and replaced with 2x120GB. THese latested 18 months and then the RAID board started complaining that one disk had bad sector (sector repair). Before loosing my data, I decided to change the faulty disk with another 120GB I had in the cupboard.

I then check the 120GB disk with badbloack:

badblocks -wvs -c 10240 /dev/sdf

THe results was that there were several bad blocks so the RAID board did a good job.

I then decided to check the old 80GB disks with badblocks. They are both reported as OK by the software, which I find unusual. SO the question is:

who is likely to be making a mistake with the 80GB disks? Badblocks or the RAID board? Maybe I could have kept going with my old 80GB RAID 1 configuration and could have saved the money I spent for 2x120GB disks?

Regards
Antonio

Both are likely to be right. It is possible to recover from bad blocks. The hard drive has spare blocks beyond the end of the drive. The hard drive firmware will assign spares to bad blocks and you can keep on running. When all of the hardware spares are used up then some file systems will work around bad blocks when you format a partition. As an experiment you could format a partition using mkfs and the -cv option and see what it tells you about bad blocks.

----------------
Steve Stites

P.S. Bad blocks are a hardware error and like any hardware error bad blocks can be intermittent. Sometimes they are good and sometimes they are bad.

NOt really following you now. Maybe I forgot to mention that the only thing I did with the harddrives before running badblocks, was to wipe out the partition table, and create one huge LInux 83 partition but without creating a file system on it.

The -c option calls badblocks doesn't it? SO what is so different from what I did?
I would have thought that running "badblocks -vsw -c 10240" would be the best test I could get with the most reliable results.

Based on what you are saying, the 120GB disk that showed huge amounts of bad blocks, had run out of spares.
Instead, the 80GB disk, even though it probably has Bad blocks (that were identified as such by the RAID board) they were remapped to spares by the drive and therefore, cause the drive still has spares, the "badblocks" software doesn't see any bad blocks?

Cheers
Antonio

Nothing really. But you would get an idea of how many bad blocks mkfs successfully works around.

---------------------
Steve Stites