TXTSETUP.OEM and CD-ROM?

  • Thread starter Thread starter Gerry Hickman
  • Start date Start date
G

Gerry Hickman

Hi,

Trying to make an unattended CD-ROM with SPs for use with some
unformatted Dell servers; Perc 3/Di RAID controller.

I made the bootable CD with integrated service pack and winnt.sif in the
i386 folder. The winnt.sif contains sections for [MassStorageDrivers]
and [OEMBootFiles]. So far so good.

I followed Oli's advice from http://www.willowhayes.co.uk/ (Great site!)
and placed the $OEM$ folder in the root of the CD instead of under i386;
I have no idea WHY it has to go in the root, but I'm sure there's a good
reason.

I then created $OEM$\Textmode, and placed TXTSETUP.OEM in there together
with all the *.INF and *.SYS for the controller.

Obviously it does not work:)

The CD boots up, loads about 100 ancient SCSI drivers then says

"TXTSETUP.OEM caused an unexpected error (18) at line 1044 in
D:\NT\Private\NTOS\BOOT\Setup\OEMDISK.C"

LOL! There's only about 20 lines in my TXTSETUP.OEM (I guess it merged
it with TXTSETUP.SIF or something?) But where did this "D" drive come
from?? There's no logical drives yet! It's a new server with RAID sets
but that's all.

All we want to do is get the driver into memory so we can manually
choose a partition in textmode portion of setup and then let the rest
run automated.

About 100 possible issues spring to mind at this point:

1. The TXTSETUP.OEM I'm using was designed for a floppy drive, and
although I included the "tag" file in OEMBootFiles, it probably gets
confused.

2. Even if it could see the "tag" file, it probably would not be able to
find the driver files, because they're in $OEM$\TextMode (which resides
in an "unsupported" location at the root of the CD)!

3. Syntax for [disks] is different for NTFS and FAT, but what about
before you've got either!

So I'm assuming there must be a simple solution? Next step looks like
hacking TXTSETUP.SIF, horrible.

There seems very little into on the net about this kind of CD, apart
from a few people saying they couldn't get it working. It seems like a
really good thing to have? Network setups are nice for clients, but if
my main server is broken, it's not much use having a network
distribution point sitting on it! Or if the network was broken, or
over-run with viruses a CD could again be useful to get up and running
in a quick and safe way.

I read Patrick's TXTSETUP.OEM "story" at

http://www.mail-archive.com/[email protected]/msg00107.html

which was very helpful, but I think is more related to a "network"
installs. Same issues though.
 
Hi, Jerry. I wrote down everything I know about this here:

http://groups.google.com/groups?selm=s5gvfmqhpbx.fsf@patl=users.sf.net

In short, I know of nobody who has gotten TXTSETUP.OEM working from
CD-ROM. Wouldn't you rather install from the network anyway :-) ?

That said, some people claim to have gotten something to work by
hacking directly on i386\hivesys.inf and i386\txtsetup.sif. Your best
bet is to ask over at msfn.org:

http://www.google.com/search?q=txtsetup.oem+site:msfn.org&filter=0

Good luck!

- Pat
 
Hi Patrick,
In short, I know of nobody who has gotten TXTSETUP.OEM working from
CD-ROM.

Your're right, it looks like TXTSETUP.SIF is the way to go. I can only
praise the person at Microsoft who (back in the days of NT) decided to
use a pure text file for setup paramaters.
Wouldn't you rather install from the network anyway :-) ?

I'm open to debate... Obviously I use the network for client roll-outs,
but for servers I'm not so sure. Here's a few reasons:

1. The server you're trying to re-build is the one that's got the
distribution point on it... err _had_ the distribution point on it
before it blew up.

2. Some kind of disaster has occurred and your ethernet switch is hosed,
you don't have a network, but you need to start building the server so
it's ready on Monday as soon as you plug in the new Ethernet swtich
that's on it's way by courier. *Or* it's a week-end and the network guys
have decided to do maintenance all week-end, sorry no network today.

3. You have massive RAID sets a 3rd party controller and with NTFS
partitions. Your O/S is trashed by a virus. The O/S backup is no good.
You want to re-install the O/S to unpartitioned space, but don't want to
trash all the other data partitions. With CD-ROM you just boot up and
choose "Create new partition", format to NTFS and off you go... with the
network, how can you do this?
 
Gerry Hickman said:
I'm open to debate... Obviously I use the network for client
roll-outs, but for servers I'm not so sure. Here's a few reasons:

1. The server you're trying to re-build is the one that's got the
distribution point on it... err _had_ the distribution point on it
before it blew up.

Well, I use Linux for most of my servers. They don't blow up.

(Half-serious. My most heavily-loaded Linux server, which provides
file, print, and Email services for ~100 users, has not been rebooted
for 206 days. It went down 206 days ago for electrical work in the
building.)

But to say on-topic... Your installation CD could fail, too. You
need to have a copy somewhere; it may as well be on another machine.
2. Some kind of disaster has occurred and your ethernet switch is
hosed, you don't have a network, but you need to start building the
server so it's ready on Monday as soon as you plug in the new Ethernet
swtich that's on it's way by courier. *Or* it's a week-end and the
network guys have decided to do maintenance all week-end, sorry no
network today.

I think your notions of "server" and "network" are too narrow. A
laptop can be a server. A crossover cable can be a network. In fact,
one of my users deploys systems at customer sites this way:

http://www.mail-archive.com/[email protected]/msg00521.html

Personally, I would use Linux on that laptop (to get DHCP etc. for
free), but he uses XP.
3. You have massive RAID sets a 3rd party controller and with NTFS
partitions. Your O/S is trashed by a virus. The O/S backup is no
good. You want to re-install the O/S to unpartitioned space, but don't
want to trash all the other data partitions. With CD-ROM you just boot
up and choose "Create new partition", format to NTFS and off you
go... with the network, how can you do this?

With <http://unattended.sourceforge.net/>, you tell it you want to
partition manually, and then you use fdisk to create a partition and
mark it "active". That's it. A little clunky, perhaps, but this is
not exactly an everyday scenario.

And we have not even started discussing the advantages of network
installs... But you know about them already.

I also have a user who has taken Unattended and made it all work from
bootable DVD-ROM. So he can reinstall the OS, hotfixes, and
applications without needing a network. We will be merging his work
into the mainline eventually since other people seem to want it,
although I cannot imagine why :-).

- Pat
http://unattended.sourceforge.net/
 
Hi Patrick,
Well, I use Linux for most of my servers. They don't blow up.

Hehe! I agree that Unix/Linux can be a nice O/S for certain tasks, but
(to me) much of it is still stuck in the stone age. Just look at Apache
vs IIS or .NET vs PHP or SQL server vs Postgres/MySQL. Then there's the
problem of integrated authentication, impersonation etc. In theory I
should be able to run my file server on Linux instead of Windows, but
with tens of thousands of NTFS permissions authenticating against a mix
of trusted NT and AD domains, how will my Linux box handle this? Surely
not Samba? What about the NTFS permissions on the backup tapes, can they
restore to a Linux box?

It could be argued that Microsoft has deliberately damaged interop in
order to tie people to it's products (which I'd agree with in most
cases) but that doesn't help solve the problem...

The other problem is that companies like Red Hat are becoming just as
bad as any other corporate. I'd be thinking very carefully before giving
them any money; dealings I've had with Sun have also been a disaster. On
the whole Microsoft have been a reasonable company to deal with,
especially considering the size, but evaluating XP has caused me to lose
confidence in Microsoft's roadmap in the context of the enterprise and
that's why I'm interested in alternatives.
(Half-serious. My most heavily-loaded Linux server, which provides
file, print, and Email services for ~100 users,

So do they use IMAP? What client do they use?
has not been rebooted
for 206 days.

I don't reboot mine very often; it's usually if I have to apply a patch.
But to say on-topic... Your installation CD could fail, too. You
need to have a copy somewhere; it may as well be on another machine.

I'm not sure it's a reasonable comparison. Burning an ISO to take home
is fairly straightforward.
I think your notions of "server" and "network" are too narrow. A
laptop can be a server. A crossover cable can be a network.

I agree it's an option, but I can't stand laptops and you have to keep
them patched. I could (of course) run a cross-over cable between two
servers that are in the same room, but it's messy. Private TCP/IP
addresses in DOS is not what I want to be battling with in an emergency
situation.
With <http://unattended.sourceforge.net/>, you tell it you want to
partition manually, and then you use fdisk to create a partition and
mark it "active". That's it. A little clunky, perhaps, but this is
not exactly an everyday scenario.

Does it support huge RAID based NTFS partitions on 3rd party
controllers? If it does, it sounds good, but this is all "built-in" when
you use a CD-ROM, you can partition, format, choose a partition and all
in an easy to use text based screen. Handy if you want to restore to
different hardware - you may suddenly have hard drives of different
sizes, different RAID sets etc., when you buy a new server. This is
where some types of scripted partitioning schemes fall down.
I also have a user who has taken Unattended and made it all work from
bootable DVD-ROM. So he can reinstall the OS, hotfixes, and
applications without needing a network.

That's what I mean, but you can do this on CD too.
 
Gerry Hickman said:
Hi Patrick,


In theory I should be able to run my file server on Linux instead of
Windows, but with tens of thousands of NTFS permissions
authenticating against a mix of trusted NT and AD domains, how will
my Linux box handle this? Surely not Samba?

Why not Samba?
What about the NTFS permissions on the backup tapes, can they
restore to a Linux box?

In theory, current releases of Samba will map between NT ACLs to/from
the client and POSIX ACLs on the local file system. So you can either
use a native Linux backup utility which understands POSIX ACLs, or you
can restore via a network connection.

In theory. I am not claiming I have tried it.
It could be argued that Microsoft has deliberately damaged interop in
order to tie people to it's products (which I'd agree with in most
cases) but that doesn't help solve the problem...

I do not really blame Microsoft for trying to sabotage
interoperability. They are just trying to make as much money as they
can, like any corporation is supposed to do. That part I understand.
What I do not understand is why customers tolerate it, since it only
serves to make everything more expensive in the long run.
The other problem is that companies like Red Hat are becoming just
as bad as any other corporate.

But Red Hat is not Linux, which is sort of the whole point. (This is
in contrast to Sun/Solaris and Microsoft/Windows.)

No matter how badly Red Hat would like to "own" Linux, they can't,
because the license forbids it. So you can always switch to a
different Linux vendor. And I am.
So do they use IMAP? What client do they use?

Yes, most use IMAP and either Outlook Express or Mozilla. Some use
random clients which you probably have never heard of (like Gnus).

The sales droids who need "shared offline calendars yadda yadda yadda"
are on an Exchange server. It seems to randomly need a reboot every
now and then. Their inbound and outbound Internet mail still go via
the Linux box, though, for security.
I don't reboot mine very often; it's usually if I have to apply a patch.

I apply patches without rebooting simply by restarting the affected
service. The mail service (MTA) is qmail, though, so there are never
any patches to apply... The last official release of qmail was almost
six years ago. (This is what we call "stable software".) Nobody has
ever found a security hole in qmail.
I'm not sure it's a reasonable comparison. Burning an ISO to take home
is fairly straightforward.

So is dragging and dropping an installation folder :-).
I agree it's an option, but I can't stand laptops and you have to
keep them patched. I could (of course) run a cross-over cable
between two servers that are in the same room, but it's
messy. Private TCP/IP addresses in DOS is not what I want to be
battling with in an emergency situation.

You would build it and test it once, just like the CD-ROM.

And you have to keep your CD-ROM patched too, don't you?
Does it support huge RAID based NTFS partitions on 3rd party
controllers?

Assuming hardware RAID, yes. A RAID controller will implement the
relevant BIOS calls which let DOS talk to it, no matter how modern it
is. You have to do the FAT->NTFS conversion thing, but that happens
automatically.
If it does, it sounds good, but this is all "built-in" when you use
a CD-ROM, you can partition, format, choose a partition and all in
an easy to use text based screen. Handy if you want to restore to
different hardware - you may suddenly have hard drives of different
sizes, different RAID sets etc., when you buy a new server. This is
where some types of scripted partitioning schemes fall down.

FreeDOS fdisk supports the notion of "use X% of the remaining space on
drive N" for arbitrary X and N, which is pretty useful for scripting
partitioning schemes in a drive-size-agnostic way.
That's what I mean, but you can do this on CD too.

Depends on the size of your applications, I would think :-).

- Pat
 
Back
Top