Diskless install of Woody
I've got back to having a further look at this, and I've decide one
reason it's not documented is that it doesn't actually work.
In order to install off the network, the procedure should go like this:
1. Configure DHCP server
2. Configure tftp server
3. Position a kernel and initial ramdisk somewhere, maybe in /tftpboot
4. Configure pxelinux.
Now, I have 1 and 2 done, and I can boot RHL's installer, DRDOS and
FREEDOS, so I figure that part's fine.
What Debian should have and does not, that I've discovered, is a menu
for pxelinux, and a selection of kernels and initial ramdisks. I think I
can get them by picking apart one or other of the various linux.bin and
root.bin images, but the ones I've looked ar are odd.
What I expected to see is syslinux configuration files saying something
like
label debian
kernel vmlinux
append initrd=installer ....
Instead, every single append statement says "noinitrd," and (at least in
the image I looked at), there were many duplicate stanzas.
That said, I know that the installer works off CD, so I figure there are
still things for me to learn.
In Red Hat Linux, at least to 7.3, you typically need one 1.44 Mbyte
floppy. It contains a modularised kernel, syslinux and its files, and an
initial ram disk.
That initrd has enough (we hope) modules in it to find typical SCSI and
IDE disk drives (including CDs) and a small selection of network modules
- 8139too, Intel 10/100 and some others.
There is room on the floppy for a kickstart file for automated installs,
and advanced user can fiddle round customising the modules on the floppy
(they're in a gzipped cpio archive). Ordinary users can simply insert a
drivers disk when needed, but for common cases only the one floppy is
needed.
Once the installer has configured the CD, HD drive or network, it gets
the main installer, equivalent to the root.bin floppy.
Anyone who wants a closer look can download the 1.44 Mbyte floppy from
RH and take a look;-).
Okay, I've said my piece. Well, most of it;-)
I will say at this point that I find some of the terminology confusing.
The so-called rescue floppy does, I predict, rarely get used for
rescuing. However, everyone does use it to boot the installer (even if
they boot from CD).
RHL has a facility to create rescue disks. They're built using an
installed kernel and (by default) modules configured to be used in
/etc/modules.conf. They can be used to perform a standard boot of Linux,
just by inserting it in fd0, turning on the computer and waiting for
events to take their course. One computer here I boot that wall all the
time as the BIOS doesn't recognise the 40 Gbyte drive, and my daughter
boots hers that way because NT (on her first drive) is being difficult.
Unless it's the drive.
So when I read "rescue," I think "is this a documentation error? How
should I interpret this?"
Phew!.
>From the Debian perspective, is it worth continuing with getting the
diskless install working using the old installer? If so, what source of
kernel should I use, and is the installer likely to work (I think it is,
but more informed opinions are welcome)?
--
Cheers
John Summerfield
Please, no off-list mail at all at all. This address accepts mail only from Debian lists.
Reply to: