FreeBSD networked install

Eugene M. Zheganin emz at norman-vivat.ru
Sat Mar 15 06:54:03 PDT 2008


Hi.

I want to ask if there will be some changes in the FreeBSD networked 
install procedure.
The situation really needs to improve.

Since the absence of a CDROM drive in a modern server environment seems 
to become common (and the legacy floppies are already dead), I think 
it's vital to have some other, non-CDROM methods of installing. Working 
methods.

For example, during last halfyear I had 2 servers with sata-only drives, 
and without legacy IDE power jacks.
The only methods of installing FreeBSD on these servers were:
- to extract a hard drive and to install elsewhere
- to get the SATA CDROM
- to copy the 6.2 installation CD onto read-write fs and distribute it 
over the TFTP with the help of pxeboot (this method is unsuitable for 
newbies, because the sysinstall doesn't start automatically, because 
read-write fs really required, but mostly because it's undocumented now).

I have and USB-CDROM, but 
http://www.freebsd.org/cgi/query-pr.cgi?pr=i386/115854 prevents me from 
using it (it really does crash on both servers).
The same PR blocks me from using USB-Floppy.
I tried to do the same TFTP trick using 7.0-RELEASE CD, but after 
loading kernel it says 'No init found' and panics. :/

Before the 5.3-RELEASE there was a great thing - a 'boot.flp' image. It 
was a fake 'floppy' image of 4MB. It was possible to extract its 
contents into the TFTP root and boot the FreeBSD via TFTP with no pain.
After the 5.3-RELEASE the naming scheme of floppy images was reset, and 
'boot.flp' became 1st booting floppy, but the real 'boot.flp' was 
completely wiped. Most of the googled howtos are built on this method, 
and I really think it was great and quick.

I think this method needs to be restored. Or at least the CD/TFTP method 
needs to be fixed to stop requiring read-write NFS and use the mfsroot.

Eugene.


More information about the freebsd-current mailing list