Very Computer

The following message appeared on the console while running bru(1) backups:

  dks1d4s7: media error: unrecovered data block read error, (data byte 18).
            Block #4410(9687).

It seems to me that fx(1M) should be used to fix this bad block.  Is there
anyway of figuring out which file is located at "Block #4410(9687)"?  

I remember sparing a bad block quite some time ago and there where two options
for adding a bad block (addbb).  For performance reasons should the bad block
be spared (replaced with another block elsewhere on the disk) or omitted (not
replaced with another block elsewhere on the disk)?

When adding a bad block (addbb), is it necessary to backup the disk, add the
bad block (using addbb), remake the filesystem (using mkfs), and then restore
the disk?  Would a simple alternative be to fsck(1M) the disk?

Thank-you for your time,

Systems Programmer                    Phone : (519)253-4232 ext.3252
Department of Computing Services      Fax   : (519)973-7083
University of Windsor
--

Systems Programmer                    Phone : (519)253-4232 ext.3252
Department of Computing Services      Fax   : (519)973-7083
University of Windsor

| The following message appeared on the console while running bru(1) backups:
|
|   dks1d4s7: media error: unrecovered data block read error, (data byte 18).
|             Block #4410(9687).
|
| It seems to me that fx(1M) should be used to fix this bad block.  Is there
| anyway of figuring out which file is located at "Block #4410(9687)"?  

give fx block 9687.  The block in the filesystem on dks1d4s7 is #4410;
we provide no programs to identify what is using that block, unfortunately.
We have some tools inhouse to identify them, and I've pushed for cleaning
them up and shipping at least one of them, but we haven't done it yet; they
need some work to be shippable.

| I remember sparing a bad block quite some time ago and there where two options
| for adding a bad block (addbb).  For performance reasons should the bad block
| be spared (replaced with another block elsewhere on the disk) or omitted (not
| replaced with another block elsewhere on the disk)?

For SCSI, you have no choice.  All you can do is forward the block, and
the drive itself does the remapping.  On most SCSI drives that are high
performance these days (all the ones we ship now) you will get a
replacement on the same cylinder (no seek), and may well get sector
slipping, so it costs no head switch either, just one sector time.

| When adding a bad block (addbb), is it necessary to backup the disk, add the
| bad block (using addbb), remake the filesystem (using mkfs), and then restore
| the disk?  Would a simple alternative be to fsck(1M) the disk?

It's a good idea to backup the disk, just to be paranoid, but I certainly
wouldn't mkfs and restore it, except in extreme cases.  Definitely fsck
the disk afterwards if it has a filesystem (as yours does).