I'm rebuilding my RAID mirror, but the rebuild keeps dying. I suspect
this is due to unreadable files on the source disk. The last time I
dealt with this, a few years ago, I think I used dump or tar to scrub
each partition and look for read errors. Then I could delete or
refresh the troublesome files, and fill up all the empty space with
crap, which would force the disk to use a back up block whenever it
found a write error. After all, the disk only reallocated on writes,
right?

I was wondering if there was a better way to "refresh" a file system,
looking for bad blocks/read errors and fixing them.

It also seems that you'd never find a disk without ANY read errors,
particulary as disks get bigger! Surely there is a good, clever way to
manage this problem when you are rebuilding a mirror? Is there
typically a way to command the rebuild to ignore some errors?

BTW this is an old system, 4.10, and a 3ware 7000 RAID card (twe
driver).