> I'm working with a 1.1 Ghz Celeron on the client, a 500 Mhz Pentium on the
> server, and a 10/100 ethernet.  I finally was able to get my system (the
one
> without floppies or CD-ROMs) to do a netboot from a USB Zip disk.  It
found
> the server right away and started downloading the kernel.  That was at 4
am.
> It's now noon, and it's still downloading.  The line reads:
>
> "Loading 192.168.100.1:/tftpboot/kernel"  and there's a row of dots after
that
> that grows gradually.  Right now there's a row of 16 dots.
>
> 8 hours seems a bit long to do a netboot....
>
> How can I speed this up?

Actually--and you probably realized this--you're not downloading the kernel.
Most likely yout tftp service on the server is not running or not
configured, and those dots are merely indicating repeated attempts to
contact a tftp server.  A kernel download takes, literally, no more than two
seconds.  It's a 600kb file!

You have to check your configuration.  Either tftpd is not running, is
misconfigured, or (most likely) your firewall is blocking tftp services.  A
standard firewall will rightly block the tftp port, but LTSP needs are
unique.

> Also: This is an ASUS TUSI-M mobo, with an SIS900 NIC chip.  I understand
this
> board can boot using PXE.  Is this the right process to add PXE?
>
> 1) Copy BIOS image using AFLASH.EXE (and back it up).
> 2) Download PXE ROM image from ROM-O-MATIC
> 3) Use CBROM or AwardMod to add the PXE image to the original BIOS image
> 4) Use AFLASH.EXE to flash the board BIOS with the new BIOS image w/ PXE

PXE was designed to ease care and feeding of terminals that boot remotely.
My understanding of PXE (and it may be mistaken) is that it's an execution
environment present in ROM on a network card, and that it understands enough
about DHCP and TFTP to download its own PXE-compliant image and run it, and
that all configuration of the PXE bootroms is done by a config file on the
server.  In other words, there's no boot prom involved, and no kernel/boot
image is ever stored anywhere on a PXE client.  Thus you don't add it to the
bios.  At least that's my understanding, but I don't use PXE.  Although some
evidence in support of this understanding may be that many PXE NICs do not
have boot prom sockets.

Thus what you need to do is put your PXE rom image on your SERVER, and
somehow set up that server to respond to requests from PXE clients by
handing them that PXE rom image.  The pxe rom image in turn will be loaded
into memory on the client, and then will take over and start looking for a
kernel, like normal.

However, what you describe might also work (I don't know), but I wouldn't
count on it.  It's much easier to set things up the standard way, anyway.

Check
http://people.redhat.com/alikins/ltsp/pxe/pxe-ltsp-recipe.txt
There are additonal files in
http://people.redhat.com/alikins/ltsp/pxe/
that it seems you'll need.

It looks like pxe will take a bit of work to set up, but it'll be a lot less
work (and less violent) than reflashing all your bioses!  I think the whole
point of PXE in the first place was to avoid having to burn EEPROMS for all
your network cards (a problem when you have hundreds of clients) or flash
crazy bioses.
--
Francis Avila




-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_____________________________________________________________________
Ltsp-discuss mailing list.   To un-subscribe, or change prefs, goto:
      https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help,   try #ltsp channel on irc.openprojects.net

Reply via email to