Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!hp-pcd!hpfcso!hpfcdj!kinsell
From: kinsell@hpfcdj.HP.COM (Dave Kinsell)
Newsgroups: comp.sys.hp
Subject: Re: Format utility (or disk analysis)?
Message-ID: <17330033@hpfcdj.HP.COM>
Date: 17 Apr 91 21:21:35 GMT
References: <1991Apr16.142221.2923@dartvax.dartmouth.edu>
Organization: Hewlett Packard -- Fort Collins, CO
Lines: 31


>What does one do if he/she suspects that a portion
>of his/her hard disk has developed some bad sectors?

Depends somewhat on whether you have HP-IB or SCSI disks, but 
there are also similarities.  On a running system, you should check
msgbuf with the dmesg(1m) command.  I/O errors get logged there.
With SCSI disks, you might see messages like:

     "Unrecovered Read Error"
     "No Record Found"
     "Recovered Read Data with Target's Read Retries"
     "Recovered Read Data with Target's ECC Correction"

You may want to scan the entire disk to systematically check for bad blocks.
Using dd to copy the raw device to /dev/null with bs=64k works nicely
for this.  On a production system, you might want to scan the disks
every night to check for sectors going bad.

Specifically for HP-IB disks, there is a program
 /usr/diag/CE.utilities/CS80/exerciser that is intended for use by service
personel.  There's nothing being officially distributed for SCSI disks at
the moment, but I could supply an exerciser if you'd like to use it.  It
does read/write tests of the volume, and can spare single bad sectors.

Regards
Dave Kinsell
kinsell@hpfcmb.hp.com
               ^
               |
               yes, I work for HP