On Mon, 28 Feb 2000, David Kramer wrote:
> Jason Costomiris wrote:
> >
> > On Sun, Feb 27, 2000 at 06:22:57PM -0600, Vidiot wrote:
> > : > The installer DOES note that it will completely repartition your drives,
> > : > thus killing whatever was already there. Shame on you for not using the
> > : > "Custom" option, which lets you choose what to do with your drives.
In my installation cookbook for newbies, I have them use CUSTOM install.
Why? Two reasons:
1) the partitioning defaults - I don't agree with the choices and it
also assumes that the user didn't have any other pre-existing
partitions, etc. One size fits NONE, IMHO!
2) the package selection defaults - It's surprising to see what gets
included and what does not with the workstation option. Again,
IMHO, ONE SIZE FITS NONE!
(A Red Hat told me recently, though, that the reason they have only the
four or five choices is that they only have a small group of people trying
(it may be larger now) to do an aweful lot of work and they just don't
have the time to try to develop more options. Maybe it would be helpful
to have a "newbie" workstation option and a more "advanced" workstation
option (more advanced - allow choice of having some network services
start up, instead of the safer, but more brain dead, way that 6.2
(telnetd and ftpd do not start up). The decision to separate user and
daemon into separate packages for telnet, ftp, etc is perhaps a good Fine.
Does anyone want to do it for them? ;-)
A custom install may seem like a LOT of ugly work to take a newbie
through, but in my experience, it's worth it. That's why I write cookbooks.
> > :
> > : That is no excuse for the installation to blindly remove partitions without
> > : double-checking with the user that it is OK to destroy that drive.
> >
> > There's nothing "blind" about it. It tells you plain as day. You didn't
> > pay enough attention, and you killed your data. You have nobody to blame
> > but yourself. It happens to everyone once. This is your turn. Next time,
> > you will be more careful.
> >
> > : RedHat's installation has to take a lot of responsibility in this fiasco.
> >
> > Maybe they could put in a "Hey, did you really read what I just said? I'm
> > about to kill all of your data. You sure about this?" message, but that
> > smacks of Microsoft's "Are you sure? Are you really sure? Really and
> > truly? Sure you mean it?".
>
> I'd like to put a slightly different spin on this.
>
<SNIP>
>
> A common thread on mailing lists and newsgroups like this is "Why aren't
> more people using linux instead of windows, and what can we do to change
> that?" Well, here's a prime example. Yes, there was a message on some
> part of the screen. Was it enough? I didn't see the message so I don't
^^^^^^^^^^^^^
> know if the average less-experienced user would have noticed. I hate
> "Are you sure?" messages, but if a program is going to wipe out all
> filesystems, I would expect the message saying so to be pretty
> prominent, with directions on how tell it to work differently.
As someone with 20 years system management experience and after having
been in the trenches with users, recovering from disasters, I can
sympathize with Vidiot. I have seen others in such pain and I have
personally experienced the exhiliration of saving the day as well as the
helplessness of not being able to after several desperate days' (and
nights) work.
I agree with David concerning the warning screens: Was it enough?
Well obviously, it wasn't Vidiot-proofed! (Note: this is NOT a personal
flame - I couldn't resist the pun!!) I agree that this is one place where
the: "Are you sure, are you really, really sure?", repeated screens are
WORTH IT. Remember the assumption made about the expertise level
of the user choosing Workstation option - newbie ? ... Protect them from
themselves (they may be drunk? :-( If you are going to have this option
at all, then an extra failsafe screen, when something so permanent
will happen, is a must for the reputation of Red Hat's Linux as a "viable
Doze Desktop alternative for the Masses" to become a reality.
Concerning Vidiot's "excuse" that he had no access to online documentation,
and flaming RH for that, well, no sympathy for that. If he had chosen
to PURCHASE THE CD's, he would have gotten the WRITTEN Installation
document. As he's discovered the hard way, there is sometimes no substitute
for the written word. Remember, to ERR is human, to REALLY screw up
requires a computer (but that's why I have a job!) In the case of
system troubles, online docs may not be available. How often now do we
get a product with the docs only on a CD? How many times in the past
20 years have I had a system without the online docs readable? All too
often! THe convenience is nice, but always PRINT OUT the installation
notes/release notes while you can, because you never know when you'll
need them. Besides, it's very important to read release notes before
doing a system upgrade. There may be major changes, things might break.
Don't ever do an upgrade without first reading the release notes. If you
don't have them available at the moment, then go find them somewhere first.
Any computer with a web browser and Internet connection would have worked
in this case to read the installation guide online somewhere on the Net, or
someone in the area has printed docs.
I also don't buy into the excuse that it wasn't his fault that he didn't
have backups. I don't care what OS you are running, you have to do
backups to different media/locations and verify that those backups are
actually readable. When you have a production system, with years of
precious data, then you MUST treat it as precious and not take short
cuts or you have no one to blame but yourself. Remember it doesn't
matter WHAT OS you are running. A disk drive can just DIE without
warning, or worse, it can corrupt stuff for quite awhile without you
necessarily knowing it. If you have been running a system
long enough to have years worth of data, you should have learned that by
now. Don't just assume that a backup has worked before you do something
major like a system upgrade, re-/install, or whatever. I just spent
a week, including a couple of all-nighters, desperately trying to
recover 20 years worth of data for a department with Oracle on a Solaris
box. Foolishly, the IT director (they have no Sys Admin currently),
commissioned a company to re-engineer (he thought for a simple tune-up,
the consultant thought otherwise) his precious database a week before they
had a major production deadline (registrations for their summer programs).
The almost-fatal assumption by the IT director: they had good backups
made just before the changes were made (moving database files across
four different partitions!). Well, there was a backup script run by
cron every Sunday night, but it hadn't run successfully for quite
awhile; no one had noticed. They had a lot of work done on Y2K fixes of
their data and structures, and backup had supposedly been done then,
too. What I found was that I had to go quite awhile back to find a
complete backup set that actually worked and was still readable. It was
the last chance, because they had re-used tape sets instead of keeping a set
once a month for a whole year, and one set per year in perpetuity. Again,
my 20 years ofsystem management experience on a variety of systems says:
VERIFY the full backups. Even make a specify one and VERIFY it before a
system change. Despite the fact that, as I recall, Vidiot was trying to
recover from a problem and so it wasn't by choice that he was attempting to
re-install, re-installation is still a major change and shouldn't be
undertaken without known working backups. I'm a system veteran and
I don't trust myself to never screw up when it comes to anything so
permanent as partitions and re-installation. Recovery and boot floppies can
be invaluable in case of emergency, so that partition-by-partition backups
can be made first, even in the worst of times.
In Vidiot's case, since he knew that his tape drive wasn't working, then
he was gambling that nothing would happen. Heh, it's electronic/magnetic
medium. Volatile stuff. FTP to somewhere else or borrow a tape drive.
Don't just back up to one place either! Verify that it actually worked -
that you can actually read what is written. Backing up to another disk
drive is a good idea, but don't rely on ONE backup. What happens if you
have a fire? Off-site storage of a backup tape or zip disks or whatever
if the data is precious enough. There are reasons why such proceedures
are recommended. Stuff actually happens!
What data do you NOT need to religiously back up on a production system?
Only the stuff you can afford to lose. What data can you afford to lose?
Not even Microsoft and "Idiot Unix" can protect people from not doing enough
backups. These are computers, and we take their reliability for granted.
I would venture to guess, though, that most people in this world are in
the same place with regard to backups - they don't really have any.
Happens all the time with students at a university. A floppy goes bad
and suddenly, the paper due in 1/2 hour is gone, or a thesis, the grad
student's whole life, disappears. The big lesson: don't take magnetic
media's reliability for granted - do backups - to multiple locations/media.
***************************************************************************
Jerry Winegarden OIT/Technical Support Duke University
[EMAIL PROTECTED] http://www-jerry.oit.duke.edu
***************************************************************************
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
