SUMMARY: format oddities

From: Dave Mitchell (
Date: Fri Feb 26 1993 - 02:13:37 CST

Hi sun managers.
This is a summary of 2 questions I fired off in quick succession a week ago

1) Why is it that a disk with demonstrable errors gets given a clean bill
   of health by format>analyse>read?

2) is reformatting disk with incorrectly specified geometry dangerous?

I got several replies, and I also spoke to Sun directly.

1) Some people kindly explained to me why I might have bad blocks, or
how to fix them; however, this wasnt was I asked - I just wanted to know
why format wasnt picking up on them in the first place. Anyway,....

The read test in format is _USELESS_ - dont bother to use it. If it
gets soft errors during reading, it will retry and if it eventually
succeeds, will continue >>>>>>without so much as a warning
message<<<<<<. Its only real use is as a basic sanity check, eg to
check that the disk geometry isnt so far out as to reference illegal
block nos.

Sun recommend the write/read/compare test - apparently this _will_ report
soft errors, and is the "strongest" test available.
Also, the guy at sun recommended letting the drive warm up for 1/2H or
so before formatting, to avoid alignment problems.

2) some people pointed out that specifying the wrong geometry may
degrade performance; I wasnt bothered by this, I just wanted know whether
it was dangerous, mainly due to the manufacturer's defect list getting

One person told me that defect lists were actually stored by block number;
format then displays them as cyl/hd/sec, in which case it should be okay.

One person thought reformatting was a good idea as long as the drive supports

                        * Personal comment *

My basic feeling after having posted several SCSI-related queries over the
months/years is that:

1) most people, including me, are just plain confused by it all
2) there are no definitive documents on the subject anywhere,
3) most knowlege out there consists of folklore and personal experience,
4) the folklore is confused and contradictory,
5) and that someone somewhere sometime ought to write a book on the subject,
   and make a _lot_ of money!

Finally, thanks 2

Geoff Halprin <Geoff.Halprin@au.oz.fulcrum> (Joshua Walfish SunOs/Macintosh/PC LAN Administrator 212-648-4473) (Matt Goheen)
glenn%upstage%ups@COM.Sun.Aus.fourx (Glenn Satchell - Uniq Professional ServicesJoel Abbott <>
Mike Raffety <>
peterg@com.murphy (Peter Gutmann)
Kevin Cosgrove 642-2676 <>
dan@com.bellcore (Daniel Strick)
mike%trdlnk@NET.UU.uunet (Michael Sullivan)
Bob Hendley <>


* David Mitchell, Systems Administrator, email:
* Dept. Computer Science, Sheffield Uni. phone: +44 742-825573
* 211 Portobello St, Sheffield S1 4DP, UK. fax: +44 742-780972
* Standards (n). Battle insignia or tribal totems

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:30 CDT