The Big Disk

Copyright Dr Alan Solomon (1986-1995)

"You're not going to believe this one", said Johann.  Well, I do see a
lot of curious cases, but I have to admit, Johann is well in the
running for the biscuit.

He started off with a nice little NEC 70 meg disk drive.  It wasn't
RLL certified, but he put an RLL controller on it anyway, because that
gives you about 50% more disk space.  I'm not sure whether RLL is a
good idea or not, but it definitely isn't a good idea to do what
Johann did.  In his defence, I don't think he knew a drive had to be
RLL certified.

That gave him a 110 megabyte drive, and DOS can only handle 32 meg, so
he used one of the systems that lets you format a drive to any size
you like.  He used Speedstore, and he had three partitions.  The first
was a small 1/2 meg partition, drive C, which was purely a boot
partition.  That had 6 cylinders, 10 heads and 17 sectors per track.
Next he had 40 cylinders which was formatted to 26 sectors per track
(the RLL standard) which gave him a 5 meg drive D.  That had all his
software - dBase, C, Turbo and so on.  Then he had a monster 100 mb
partition (drive E), which he used for data.  I don't like partitions
greater than 32 meg, so I asked him why he didn't just use four 32 meg
partitions.  He explained that he had 20 meg databases, and how can
you do serious work on that in a 32 meg partition?  A very valid
point.

After he'd been using the disk for a while, he started to get
occasional read errors.  This got more and more annoying, and also the
files on the disk were starting to get fragmented.  So he bought a
product whose name I shan't mention, because there are several around,
and I think most of them would have given the same problem.  First he
ran the diagnostic program, which told him that he had a lot of bad
clusters - about 1 meg.  That's bad, very bad indeed, and shows you
what happens when you use a non-certified RLL disk with an RLL
controller.

So next he ran the program that was supposed to fix his disk problems
by moving data to good clusters from dubious ones, and marking any bad
clusters as unusable.  Next he ran the program that is supposed to
defragment the hard disk.  He'd run this bit before, and he said that
it speeded things up by 10 to 15%, but this was the first time he'd
run it after the fix program.

The program finished.  But when he tried to use his disk, it wouldn't
work.  It booted fine on drive C, but drive E, the big one, wouldn't
read - it kept reporting errors in reading one particular
subdirectory.  So he used the disk snooping program to have a look,
and sure enough, parts of that subdirectory were now unreadable.

I explained what had happened to him.  The fixing program had marked
some bad bits as unusable;  maybe the right ones or maybe the wrong
ones.  The defragmenting program had either tried to write the
subdirectory onto a bad bit, thinking it was good, or else was
pointing to the wrong part of disk as this subdirectory, and that was
a bad bit of disk.  In any case, it was clear that the combination of
a 100 mb partition and the Speedstore software had confused the disk
defragmenter.  So he ran the disk repair program again, and when it
had finished, he couldn't read anything on the big disk, as DOS
reported that the FAT was bad.  The FAT is the part of disk that
stores the information about what bit of disk belongs to what file -
rather vital.

"Um", he said.  At this point, he was a worried man, so he spoke to
his dealer and to the defragmenter people.  Between them, they came up
with a brilliant plan, and when he explained it to me, I was
breathless with amazement at the sheer ingenuity of it.  And aghast at
the thought of what the effect would be.

First he repartitioned his disk, so that the cylinder which gave the
bad FAT (cylinder 46) was in a partition all by itself, called drive
E, and there now was a drive F containing all the data from his big
disk, but of course no FAT or directory.  Next, he formatted this new
drive F, and then (and this is the really clever bit) he ran the
unformatting program that came with this kit of disk tools.  He also
tried to format E, but it wouldn't, because of the bad sectors.

Unfortunately, this particular unformatter only works properly when
your files are fairly small (like 2K or so), and Johann was left with
18 subdirectories with names like SUB001, SUB002, and inside those
subdirectories there was absolutely nothing worth having.

So he phoned his dealer, the defragmenter people, the insurance
company and even his accountant.  And guess what - the accountant
turned up trumps, because it just so happened that his accountant had
lost a major database a few months ago and ...  well, you can guess
the rest.

And so it was that Johann was telling me his story of self-inflicted
damage and potential ruin.  I did ask him why there was no backup, and
he asked me how you back up a 110 mb disk containing 20 mb files?
Well, actually, Johann, you can get tape streamers, Bernoulli boxes,
cartridge hard disks - my preference is Bernoullis.

I had a look at the disk, and it was all true - you could see the
trail of damage, like a rampaging bull.  But there was one little bit
of his disk that had escaped the hurricane of destruction- track 46.
Track 46 had the FAT and directory, and as far as I could see it was
mostly intact.  Not that he hadn't tried, of course - he'd tried to
format that crucial area, but it had valiantly resisted his attempt -
well done track 46.  He suggested that we run his disk doctoring tool,
but I looked at him as if he were mad, and brought out my own.

Very very gently, I persuaded track 46 to yield up its vital FAT and
directory, and I copied it all onto a diskette.  Parts of it were
unreadable, and parts of it were corrupt, but there are ways to fix
such things, provided you have the raw material.  I copied that
precious diskette, and worked on the copy, of course, until I had
completely restored it to its former glory.  Then I sent Johann out
shopping.

I told him that he shouldn't use his disk again, since it was so
obviously going flaky.  Then we contacted a friend of mine at NEC, and
explained the situation.  What we needed, was to borrow a big disk
just like Johann's, and after a bit of persuading and a promise of a
shareware Modula 2 compiler, he agreed to lend us one.

When the disk arrived, Johann and I installed it in his AT.  We then
partitioned it just like his old one, and then I copied the entire
contents of his old disk onto the new one, including most of the bad
sectors, because there are ways of telling a disk controller to try a
bit harder to read a sector - those things are made very
conservatively and give up much too easily.  Next, I copied on the FAT
and directory that I had so carefully reconstructed, and then we had a
look at his files.

The trouble with foreigners is that they don't know how to behave.
Americans always insist on shaking hands every time you meet, but did
you know that Southern Germans go in for hugging?  No, neither did I,
and most embarrassed I was too.  Fortunately, there were no witnesses,
and I certainly shan't tell anyone.

At that point, there was still a lot of work to be done.  100
megabytes had to be backed up, because the NEC disk was only borrowed,
and had to go back.  Normally, I'd finish the job, but Johann had
shown himself so expert in running complex disk fixing software, that
I thought it would be good for him to spend a little time running the
most important disk protection software of all.  It's called Backup,
it comes free on your DOS disk, and it took him 15 hours and 300
diskettes, because I don't like moving a disk that hasn't been backed
up, because you never know.  I think hard disks are much more robust
than people think, but it is possible for something very drastic to
happen to anything that is put in a car boot.

Johann will now buy himself something that can cope with backing up a
110 mb disk, and it isn't really important what he gets, compared to
having nothing at all.  My thanks to NEC for being so helpful - their
disk would probably have been fine if it had been used in the way that
they recommended.  And the lesson that Johann learned is that no
matter how difficult or expensive it is, a hard disk must be backed
up.