RAID-1 and bad sectors?
in [Hardware]
|
Prev: LCD Monitor display confusion
Next: HD replace problems
From: Folkert Rienstra on 12 Jan 2006 19:22 <void(a)no.spam.com> wrote in message news:46v9r195r6m540q3b8ef5ikn08kdjn6380(a)4ax.com > I just used BootItNG to resize the 2 NTFS partitions on my hard disk. I was > increasing the size of the C drive, so I had to shrink the D drive and slide > it down to the end of the disk. When I did the slide operation, I told it to > slide everything including the unused areas. I ended up getting an "error > reading from hard disk" message. So I retried the slide operation, but just > told it to slide the data only. That worked, so maybe there was a bad sector > in an unused area? > > Anyways, I'm now ready to set up RAID-1 in my system. But I'm wondering if > that bad sector might cause a problem in a RAID-1 setup? > I know RAID-1 is supposed to help you if a drive fails. And there are lots of broken products on the market that fail miserably when actually taken to task. > But what if a drive doesn't totally stop working... > what if it only develops a few bad sectors? > Can RAID-1 handle that? Common sense says that it should with last decade drive technology. But OSes and drivers still basically behave the same as 10-20 years ago, so who's to say. Common sense is often lacking in RAID products. > (I'll be using the HighPoint 370 controller on my Abit KT7-RAID motherboard.) > Like what would happen if the RAID controller is writing some data, and one of > the drives is OK but the other drive has some bad sectors on it? Bad sectors should only matter on reads and corrected by data from the 'perfect' drive. > Can the RAID controller detect the bad sectors and write the data to good sectors? No, unless it reads all data back that it writes, which is very inefficient and therefor doesn't happen. > Or will it not detect the bad sectors and the 2 drives will be out-of-sync? Yes and no. Badly written data will not be detected until the next read to that data so until that time the drives will not be out of sync even though technically they are. Harddrives do not write-check on standard writes. With write caching enabled they don't wait for writes to finish even.
From: Folkert Rienstra on 12 Jan 2006 19:24
"Arno Wagner" <me(a)privacy.net> wrote in message news:41l5jjF1f0m2eU1(a)individual.net > In comp.sys.ibm.pc.hardware.storage void(a)no.spam.com wrote: > > I just used BootItNG to resize the 2 NTFS partitions on my hard > > disk. I was increasing the size of the C drive, so I had to shrink > > the D drive and slide it down to the end of the disk. When I did > > the slide operation, I told it to slide everything including the > > unused areas. I ended up getting an "error reading from hard disk" > > message. So I retried the slide operation, but just told it to > > slide the data only. That worked, so maybe there was a bad sector > > in an unused area? > > > Anyways, I'm now ready to set up RAID-1 in my system. But I'm > > wondering if that bad sector might cause a problem in a RAID-1 > > setup? I know RAID-1 is supposed help you if a drive fails. But > > what if a drive doesn't totally stop working... what if it only > > develops a few bad sectors? Can RAID-1 handle that? (I'll be using > > the HighPoint 370 controller on my Abit KT7-RAID motherboard.) Like > > what would happen if the RAID controller is writing some data, and > > one of the drives is OK but the other drive has some bad sectors on > > it? Can the RAID controller detect the bad sectors and write the > > data to good sectors? Or will it not detect the bad sectors and the > > 2 drives will be out-of-sync? > > It will detect a bad sector only if it reads from it. In that case > it will fail-over to the other disk. Many RAID controllers will > also mark the disk as bad and kick it form the array. If you > have bad sectors on eiter disk in different places that behaviour > is not too desirable, but in this case you are in trouble anyways, > since modern disks only exhibit bad sectors if they are dying or > have very serious problems. Ignore the blithering idiot. Bad power supply or overheating can cause these symptoms where the drive is not at fault. > > Now, you might be in this situation and then your best bet is > possibly RAID recovery software (have seen some, don't remember > where) or if your controller allows you to disable the kicking on > defects. Linux software RAID seems not to allow you to do that. > What I would do in such a case is to mount both partitions/drives > read-only, determine which has less bad sectors, write them > down and copy that partition/drive to a good one. Then copy > the bad sectors on the first drive from the second one. If > ithe number of bad sectros is high, then script this using, > e.g., badblocks and dd_rescue. > > When writing to a bad sector, the data will be written to both drives. > Since the data is written with checksums, > then the controller will recognize a bad sector and read it form the > other disk. Nonsense, that is what the RAID driver will do. > See also above. > > Arno |