>>> On Fri, Nov 30, 2007 at  3:12 PM, in message
<[EMAIL PROTECTED]>, Thomas Kern
<[EMAIL PROTECTED]> wrote: 
> If such a tool can be made, why not go just a bit further and mimic what
> would have been done by dasdfmt and fdisk (1 partion on the whole mdisk)?=

If you're cloning guests, this is a waste of time and machine resources.  If 
you're doing installs, it might be worthwhile, if you know 100% of the time 
that every minidisk allocated to every guest will only ever have 1 partition 
put on it, and won't be part of an LVM volume group.  If you can't say that, 
then again this is a waste of time and resources.  Let the installation 
automation do what it is supposed to do, and the machine resources used get 
charged to the proper account.

> And optionally, why not put in the correct blocks that mimic an mkfs?

This is really not a good idea.  For repeatability purposes, you always want to 
do a mkfs on all file systems during a fresh install.  If you're going to be 
using LVM, you also want to do a dasdfmt during the install process, to get rid 
of the LVM metadata left over from the previous attempt.  Plus, I would be 
very, very skeptical that an EXT3 journal would be created correctly.  Mark 
Wheeler's methodology avoids having to worry about that.

> Then I could run several initializations on preallocated volumes while I =
> am still generating the target directory entry, granting vsw access, etc.

Personally, I think you're trying to use the wrong tool (CMS) to get around 
this problem, when it's not really necessary.  Let the Linux system format its 
own disks and file systems, and do them in parallel.


Mark Post

Reply via email to