Always check and double check

Always check and double check

Post by steve hove » Sat, 27 Jan 1996 04:00:00




> The other morning I set out to do a fairly simple project.  The
> client had an oversized DOS partition, a badly sized /u, and wanted
> to recover the DOS file space by cutting it down from 300 meg (!) to
> the 20 meg he really needed :-)

[snip]

Quote:> In retrospect, I *should* have checked the second set.  It only

[snip]

Quote:> Another thing I *should* have done is noted the start and end
> points of the existing Unix partition.  If I had done that, I

[snip]

Quote:

> I think some day I shall write a book on all the things I
> *could* have done that would have saved me grief :-)

You've made several false assumptions above..

How hard it is to recover and rebuild data is directly dependent on how
important the data is, and how mean to you the client will be if you lose
any of it, and how much time you have to be bothered by it all..

Therefore, no matter what you checked or tested, the outcome would still
be the same if its really important data, a really * inclined client,
and you have to catch a plane in an hour.

 
 
 

Always check and double check

Post by Tony Lawren » Sat, 27 Jan 1996 04:00:00


The other morning I set out to do a fairly simple project.  The
client had an oversized DOS partition, a badly sized /u, and wanted
to recover the DOS file space by cutting it down from 300 meg (!) to
the 20 meg he really needed :-)

I estimated this might take two hours, probably less.

The client makes complete system backups every night, so when I
arrived on site, I verified one tape completely and one tape
partially, so I felt pretty comfortable there.

I then noted that he had two sets of Emergency Boot floppies, so
I shut down and tested one of them.  Life was good, because I
could mount the hard drive, and read the tape.

Confident that I could now proceed with assurance, I booted up
dos and repartitioned the disk.  Not wanting to end up with
multiple Unix partitions, I blew off the existing partition,
restored the Dos stuff, and then blithely booted with the
Emergency floppies and...


I-hate-you-with-my-entire-soul errors and I said "Oooh.  Dang!
This is not good, but *fortunately* we have the second set of
Emergency disks, so..."

As you may remember from above, I had not actually tested the
second set, had I?  Heck, they were plainly marked, they had
a date that matched the first set, they had been carefully put
away with the first set.  You certainly wouldn't think that
anyone would have taken that second set, formatted them for DOS
and used them to copy some files, would you?  If someone *had*
done such a dastardly thing, you surely wouldn't think that they
would have carefully put those now useless disks back with the
other set, would you?  No, you certainly wouldn't think that
was very likely, would you?  Me either.  But that's what had
been done.

Well, three and a half hours later, everything was working again.  
Reinstall (floppies!), configure a tape drive, restore, you know the drill.
Not exactly how I had planned to spend the morning, but... I guess
it could have been worse.

In retrospect, I *should* have checked the second set.  It only
takes a few minutes, and certainly would have been worth it.
(I did try diskcopying the blown disk; no dice).

Another thing I *should* have done is noted the start and end
points of the existing Unix partition.  If I had done that, I
might have been able to just re-fdisk and get back up to
where I started.

As I always say, you can't be too careful, can you?  Floppy
disks do blow up sometime, and while it's rather amazing that
the darn thing worked once and wouldn't work ten minutes later,
the utterly amazing does happen now and then, doesn't it?

I think some day I shall write a book on all the things I
*could* have done that would have saved me grief :-)

--
Tony

 
 
 

Always check and double check

Post by Sean Embr » Sun, 28 Jan 1996 04:00:00


    Tony,

    I used to work in the * industry. I ran computers at race
parks that kept track of how much was bet, on what ticket, from what
forigen tote, ect. ect. (If you go to a track, you know what I mean.)

    Any way, I used to annoy people with the precautions I'd force them
to make. I kept on with the precautions, and saying "I'm sorry to make
you do this, but you can't be too careful."

    One day I said this to a guy that had been in the industry for many
years. He said:

    "I never met a guy that would wake up in the morning, hit himself on
the head and say, "Gee, I was too careful yesterday!", but I have met
quite a few that said, "Gee, I shoulda been more careful yesterday!"

    One thing to keep in mind, If the computer made a 100,000.00 dollar
mistake, the company we worked for would have to make it good. Betting is
like a cash machine, if you pay out too much, you can't get it back.

    If I were going to blow away a partition, I'd insist on two complete,
byte verified backups. Good thing too, I've had tapes break during the
restore!

    Well, at least nothing too bad happened!

--
Sean

 
 
 

Always check and double check

Post by Geoffrey Wel » Mon, 29 Jan 1996 04:00:00



> Another thing I *should* have done is noted the start and end
> points of the existing Unix partition.  If I had done that, I
> might have been able to just re-fdisk and get back up to
> where I started.

Can you do this?  Microsoft FDISK seems to go out of its way to make sure that
new partitions established in the space previously occupied by a perfectly
valid partition aren't useable.  That's why I carry with me a disk editor that
allows me to fiddle the partition table myself.

Quote:> I think some day I shall write a book on all the things I
> *could* have done that would have saved me grief :-)

Like staying away from computers?  Like not telling your investment adviser
that no one could possibly get rich building an operating system empire from a
BASIC interpreter?  The list goes on...

--
         Geoffrey Welsh, Senior Developer, InSystems Technologies Inc.

USENET has become the sanctuary of the Church of the Perpetually Injured Party.

 
 
 

Always check and double check

Post by Tony Lawren » Tue, 30 Jan 1996 04:00:00



: > Another thing I *should* have done is noted the start and end
: > points of the existing Unix partition.  If I had done that, I
: > might have been able to just re-fdisk and get back up to
: > where I started.

: Can you do this?  Microsoft FDISK seems to go out of its way to make sure that
: new partitions established in the space previously occupied by a perfectly
: valid partition aren't useable.  That's why I carry with me a disk editor that
: allows me to fiddle the partition table myself.

I don't know.  That's why I said "might".  In this case, nothing would
have been near this area until the Unix fdisk ran.  Whether or not that
does anything that would disturb an existing position, I don't know.

: > I think some day I shall write a book on all the things I
: > *could* have done that would have saved me grief :-)

: Like staying away from computers?  Like not telling your investment adviser
: that no one could possibly get rich building an operating system empire from a
: BASIC interpreter?  The list goes on...

Well, as I said, no one ever made millions from any of *my* opinions :-)

--
Tony

 
 
 

Always check and double check

Post by bill davids » Thu, 01 Feb 1996 04:00:00


| In retrospect, I *should* have checked the second set.  It only
| takes a few minutes, and certainly would have been worth it.
| (I did try diskcopying the blown disk; no dice).

I try to keep a backup copy of the boot disks on another system,
assuming that such is availabe, and that's one of the things my
laptop provides. I do a physical copy of the floppy with dd, and
store it in a subdir on the hard drive. Repeat with all the other
parts of the boot set, and then zip the subdir to a single file.

This gives me a copy on an independent machine. I do cross platform
backups, using my laptop (Linux), my office archive server (SVR4),
and of course lots of SCO machines. As long as you have that you can
format and make a new physical copy any old time.

Oh, I like Linux because it is easy to support odd formats, such as
MicroSoft install disks, etc.

| Another thing I *should* have done is noted the start and end
| points of the existing Unix partition.  If I had done that, I
| might have been able to just re-fdisk and get back up to
| where I started.

I wouldn't bet on that, honestly, but making a hardcopy of the
partition and fdisk info would have been useful starting data. At
least the backups were good.

| As I always say, you can't be too careful, can you?  Floppy
| disks do blow up sometime, and while it's rather amazing that
| the darn thing worked once and wouldn't work ten minutes later,
| the utterly amazing does happen now and then, doesn't it?

That's why I back up to hard drive, and then back up the hard drive
shortly thereafter.
--

Too many improbable things have happened to me, both good and bad.
I'm living on the edge of the bell shaped curve.

 
 
 

1. Double Checking Sun's Reality Check

Recently (I think it was yesterday) I saw a post by someone wondering
about an article from Sun lambasting the running of Linux on the z800. I
found this excellent refutation by Moshe Bar who also explains how he got
Linux S/390 running on his own personal box. An interesting read just for
that bit, I think.

http://www.byte.com/documents/s=7030/byt1015006951867/0304_moshe.html

Quote:
"Sun's article also says that Linux is really designed for the PC
architecture (ostensibly, the x86) and not for other platforms such as the
mainframe's S/390. And another argument against Linux on the mainframe is
that the Linux virtual memory manager is not suited to run within a
virtual computer like z/VM because its generous allocation of buffer cache
conflicts with I/O buffers already allocated by z/VM for its guest OS.

"That is pure nonsense. In the 2.4 kernels, buffer cache utilization is
much lower than in previous versions; most stuff goes into the page cache.
Actual paging for most commercial applications is close to zero. RAM is
cheap and abundant. In many years of consulting I have yet to see a
production-duty Linux server paging in any significant amount. Add to that
z/VM's capability to restrict the RAM allocated to the guest machine and
Sun's argument becomes null and void."

2. Installing X after install Linux.

3. Want RCS co NOT to check out if nothing new checked in

4. ISP connection problems

5. Partition check order in fs/partition/check.c?

6. How can I do this? War stories needed.

7. Checking upd check sum?

8. Change in configuration files...

9. 2.3 install check script doesn't check

10. A script that will only check file if it has been modified since last check

11. cvs check out problems..cannot check out files into the repository itself.

12. Help with fsck ALWAYS running full check (Slackware 96)

13. New kernel versions always check file systems. Why?