terryl (04/03/83)
Recently, we developed a bad spot on the inode portion of our user disk,
so we dumped and restored the disk to a new disk. We were pretty lucky, in that
the bad spot was only 3 blocks long, and a find from /usr just found just a few
mail messages in /usr/spool/uucp that had not gone out yet, so we didn't lose
anything of importance. But, dump in its infinite wisdom to speed things up
reads the inode area of the disk 16 blocks at a time, and when it came to our
bad spot, even though there were only 3 bad blocks, according to dump 16 blocks
were bad. Well, the restor went fine until we ran fsck, and then we found out
that the restor actually did not go too well. Files had turned into directories
and directories had turned into files, not to mention the usual unreferenced
files and directories. Now my real question is this: how much speed does it
buy you to read the inode area of the disk 16 blocks at a time instead of 1
block at a time??? Needless to say we had to get the old disk out, tar it onto
a couple of tapes and untar it onto the new disk because we didn't trust the
restor. Comments, anyone???
Terry Laskodi
{dec,ucb}vax!teklabs!terryl (UUCP)
terryl@tektronix (CSNET)
terryl.tektronix@rand-relay (ARPA)