Questions[pxe-boot](server)

I am working on a pet project, and I am trying to figure out where I should start with achieving the following functionality:

I have a PXE boot rom that wlil do burn-in tests, my burn in test sets a flag on the PXE/DHCP server.

If a client is booting and does not have the flag, I want to offer my PXE boot rom.

If a client does have the flag, I want to just skip pxe boot.

I can do this via pxe config -- but for entire subnets. how do i manage this per MAC?

Edit: I suppose it is worth mentioning, that I can do this via linux or windows server 2008 and it really comes down to which is an easier time to implement this. I'd probably prefer to write the interface for managing the DHCP settings in .net than java or something, but I imagine the dhcpd.conf is going to be more straightforward.

Here's a weird problem.

We've got a number of devices with dual-NIC mainboards. Some are Realtek NICs, which suck. Some are Intel e1000s, which don't.

I've just noticed on 2 machines, one is an Intel NIC, one is a Realtek, that when I put the MAC address of one machine into the dhcpd.conf file on our DHCP server to get it to PXE boot the machine into a rebuild environment, initially everything is fine.

The server gets a DHCP allocation, and PXE boots into the Ubuntu preseed enviroment.

On one or two machines, it gets as far as Ubuntu's DHCP network configuration, and fails. If i pull up a busybox shell (on tty2 on the installing machine), and run ip link, I can see that the UP flag is set on the other NIC.

Here's some stuff.

  host xeon16-ghz240-gb48-node1 {
        hardware ethernet BC:AE:C5:07:1F:18;
        filename "pxelinux.0";
        next-server 192.168.123.80;
  }

That's what's in dhcpd.conf

This is what ip link on the evil machine looks like.

Only one NIC is actually connected (deliberately).

As you can see, the NIC that's in the dhcpd config, is not marked as UP, and the link that is UP, isn't the one in DHCP.

So far I've seen this on two brands of dual-NIC configuration.

Does anyone know 1) what's causing it, and b) What we can do about it?

RFC 4578 defines various machine architectures for PXE as follows:

Type   Architecture Name
----   -----------------
  0    Intel x86PC
  1    NEC/PC98
  2    EFI Itanium
  3    DEC Alpha
  4    Arc x86
  5    Intel Lean Client
  6    EFI IA32
  7    EFI BC
  8    EFI Xscale
  9    EFI x86-64

I've tested out two IBM machines (HS22 blade, x3550M3) and they both do EFI PXE boot using the 'EFI BC' architecture.

What does it stand for? When is it used? Why is IBM using it instead of 'EFI x86-64'?

I am not sure if this is just specific to my distro's packages or is a vbox limitation. So, any help would be appreciated.

Ok, so I have a network of VMs, with one VM acting as a NAT for the other VMs in the "internal" network. One of those VMs is running a DHCP & TFTP server, and I just need to boot other VMs off of this server, but all I get starting the VMs when booting from the network is "Fatal: Could not read from the boot medium ! System halted".

This is kinda killjoy since vbox is pretty easy to use and to experiment with, but I can't get around this error. Help ?

In the past I've setup a PXE server several times by making the normal dhcp server return everything needed for network booting: both the regular ip config AND bootinfo in a single dhcp setup. That way it is easy to do.

According to the documentation i've read it (wikipedia has a nice summary) should be possible to have an unmodified dhcp server (for example if the admin at your location denies access) and a separate server that ONLY returns the boot information. This is usually called the "PXE server" or "Proxy DHCP" server. In the past I've seen (not used) this pxe server (last release was 5 years ago).

Yesterday I installed Cobbler to see what it really does. Now I would like to use it (love the ease of use) but modifying the 'main' dhcpd server to get PXE is not an option.

I've not been able to find any documentation on how I can get cobbler to run using a real PXE (==proxy dhcp) server.

Can this be done with cobbler?

Can I configure the isc dhcp server to behave as a PXE server (i.e. NO ip info, only boot info)?

Or should I go for a completely different approach (if so, what is your suggestion) ?

Thanks

New discovery I just made: The changelog for dnsmasq version 2.4.8 says:

"Proxy-DHCP, where dnsmasq just supplies the PXE information and another DHCP server does address allocation, is also allowed."

VERY INTERESTING. Especially because cobbler can already control dnsmasq.

Update:

dnsmasq 2.51 will support this config that fully does the trick I wanted.

My /etc/cobbler/dnsmasq.template now looks like this:

# Cobbler generated configuration file for dnsmasq
# $date
#

# resolve.conf .. ?
#no-poll
#enable-dbus
read-ethers
addn-hosts = /var/lib/cobbler/cobbler_hosts

# Be a proxyDHCP server
dhcp-range=10.10.0.0,proxy

# Only respond to clients that are known (i.e present in /etc/ethers)
dhcp-ignore=#known

# Set this (and domain: see below) if you want to have a domain
# automatically added to simple names in a hosts-file.
expand-hosts
domain=test.basjes.nl,10.10.15.0

# Loads <tftp-root>/pxelinux.0 from dnsmasq TFTP server.
pxe-service=x86PC, "Boot PXELinux (=Cobbler controlled)", pxelinux ,$next_server

$insert_cobbler_system_definitions

Update: 2012-04-30

A few months ago I got an email from someone stating he couldn't get the above to work. Turns out that I had done some fixing and changing of my own cobbler that I had forgotten about. So I provided the crucial fix to the mainline cobbler which has just become part of the trunk. I also created some additional supporting documentation.