On Fri, Feb 23, 2007 at 14:40:34 -0500, Jack Howarth <howarth@xxxxxxxxxxxxxxxxx> wrote: > What worries me is if such a scenario would become even more > complicated if LVM were involved. FYI, it seems to me that > bad blocks are the weak point of RAID-1. It can cope fine with > drive failure but can become quite confused if one of the two > drives develops bad blocks. Not that particular one as you were able to fix things at the raid level and didn't need to care what the meaning was of the contents of the drive. I wouldn't recommend using Promise raid on consumer motherboards. Linux software raid is more flexible and may be faster as well. Linux's software raid won't get confused when running accross bad blocks on a drive. It will fail that drive out. What is bad is if you get bad blocks on more than one disk that aren't detected, than rebuilding is a lot more of a pain. You can guess that normally the content from the other drive can be used, but because the drives are not exactly in sync after one is failed out of the array, this may not always work. If you then have to work out what file might have been corrupted from knowing which blocks were bad, then LVM will throw in additional complications. P.S. Did you get a copy of the set of kernel rpms you were looking for or is the issue moot now?
