Get ethernet MAC address without powering on machine?

[Robert G. Brown]

On 4 Jun 2002, Jim Meyer wrote:

> Howdy!
> 
> We buy commodity boxes not only for our distributed computing needs (a
> render farm) but also for our desktops. We like to use kickstart to set
> up the box; we've honed the config for this pretty well. The problem is
> we need to get the MAC address prior to kickstarting the box so that we
> can give it a trusted IP which in turn allows it to NFS mount the disk
> where the RedHat distributions are.
> 
> This leads me to my question: anyone know of a clever way to get the MAC
> address from an ethernet device without needing to power the machine up,
> load an OS, and probe around a bit. Having noted that ethernet devices
> often seem to display link and traffic lights even when the computer is
> powered down, I was thinking that a fairly special purpose device might
> be designed (RJ45 plug and simple digital readout kinda thing) which led
> me to wonder if such a device already exists.
> 
> Has anyone ever seen or heard of such a thing?

We get our commodity box vendor (intrex.com) to determine the MAC
address while building and/or burning in the system and to label the box
with it/them.  Then we can just read them and put them into the
appropriate tables.

Many NICs have the address printed on the device (on the Tyan 2466
motherboards, for example, with built in 3c920, there is a label on the
RJ45 housing).  It is easy for the vendor to do, a pain for you to do.

The alternative(s) have been discussed on the list a few times fairly
recently -- check the list archives for the last five or six months.  If
your systems use PXE/DHCP to install (and if they aren't, they should
be) one can just power the system up with the bios set to boot from the
PXE netbootable device, and then read the MAC address off of e.g. the
dhcp logs.  Some folks have automagicified this process so they boot,
wait, (scripts work some magic) and boot again, with kickstart/pxe then
completing the install the second time, concluding with a boot into
production.

BTW, I wouldn't recommend doing an NFS install.  Our own experience is
that an http kickstart install is the most robust and requires that the
installing host have no special privileges.  NFS is relatively slow, is
safe only in a firewalled LAN sandbox, and requires all sorts of special
privileges.  HTTP is fast, recovers decently from network errors on a
slow network (FTP doesn't, which is why I don't recommend it), works
fine if you are on the east coast and your install coast is on the west
coast (bandwidth permitting:-) and works fine without any special
privileges even if the RPM server doesn't really "trust" the host doing
the installing (obviously the installer must trust the install host).

Installing via the web or ftp is easy.  Just create a web server with
your standard red hat directory laid out on it, and direct your
kickstart file to look there.  I can probably come up with an example
kickstart file if it would help.

   rgb

> 
> Thanks!
> 
> --j
> -- 
> Jim Meyer, Geek At Large                              purp at wildbrain.com
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu