If you’ve arrived at this blog, you likely already know what iPXE booting is, but for those who don’t, I want to give a brief introduction to what iPXE is and how useful it is; both for IT engineers and computer hobbyists.
So, What is it?
Quite simply, its an open-source bootloader, primarily for the x86/amd64 machine architecture. Its aim is to allow you to boot software and images across a network connection, using common network protocols such as tftp, http, ftp, iSCSI and AoE.
iPXE can be called over the network, started from bootable media like USB sticks or CD-Roms and even injected into the ROM of network cards where available. This makes it an ideal tool for IT ‘fixers’ who often need to load a variety of images rather than just the standard Windows install on a machine.
A Brief History of PXE Booting.
PXE (Or pixie) booting is nothing new – in fact, pretty much all of the x86 type computers I have come across built in the last 10 years allow PXE booting from the network card, although sometimes the option needs to be turned on in the BIOS or the NIC. Thin Clients, VoIP phones and other ‘low touch’ devices also often support downloading configurations, software patches and other such updates over the network from a central source. But PXE booting is not without its problems. Its not setup to allow user interaction, meaning whatever the DHCP config is set to delivered is delivered. Secondly, it uses tftp to transfer files – perfectly adequate for small text based config files, but SLLLOOOOOWWWW when sending much over a few megabytes, even on fast lan connections. Finally, PXE is only available with the functionality built into the firmware, so if it doesn’t do what you need it to, you’re pretty much stuck with what you’ve got.
iPXE has had a bit of a convoluted development path in its reasonably short life. It began life as ‘Etherboot‘, which started in around about 1995. Etherboot evolved into gPXE (Nothing to do with Google) and then after a bit of a falling out within the gPXE team, gPXE was forked by some members of that team into the iPXE we have today. Note, iPXE doesn’t have anything to do with Apple and their iDevices. Both gPXE and iPXE continue to this day, doing pretty much the same thing. However, from my experience, gPXE is better documented, whilst iPXE receives more regular development activity. And development does not happen in parallel, so you may want to check out both in case one has a feature set the other doesn’t.
A Google Tech-Talk about gPXE and Network Booting.
I’ve moved from gPXE to iPXE in 2012, mainly because gPXE wouldn’t support a NIC on a Dell machine I was trying to boot, and had been considering making the switch at some point anyway. iPXE resolved the issue for me, so its now become a bit like Doc Martins were in the 80s and 90s – The boots of choice.
But why use iPXE Booting?
For me, it was a means to an end – I needed to test a motherboard and RAM combo, but didn’t have a spare hard-disk to boot from. I also didn’t have any blank CD’s to burn a live-cd image to in order to boot from a CD drive. Finally, I didn’t have any large (>64mb) USB thumb drives that I could write an ISO to, so I was kind of out of options. Then I wondered if there was a way of using the small stick to chain boot something from the network and well, it turns out there was.
I’ve stuck with network booting because once you get it working, your life as an engineer becomes much easier. No more messing around burning CD’s, or digging through CD wallets trying to find the version of the particular utility you need.
Now I can connect up a network cable, boot from the network, choose the ISO, utility, or test tool you require from a menu, and stand back and watch it boot from my NAS filer. This is often quicker than waiting for a silver platter of plastic and metal spin up and be read using the power of laser light.
The process is a little complicated, but once you’ve set it up the booting piece is pretty much hands off; the only maintenance is adding new images and updating the menu.
The way I’ve configured it is
- PXE requests an IP Address from the DHCP Servers
- IP Address is returned, along with instructions on getting the iPXE ‘software’ from a TFTP server.
- The PXE downloads iPXE from the network, and chainloads into this.
- iPXE re-requests an IP Address from the DHCP Servers, along with iPXE specific instructions – i.e. where to go next.
- iPXE reads through this config and initiates a http connection to the web server containing the main iPXE config files and menu etc.
- The end user/operator (me) chooses what to boot from the menu.
- iPXE parses the settings supplied in the menu, downloads the necessary components (again, over http) and boots the image.
gPXE took me about a day to get working, because of the need to install and configure a DHCP server with various options. Because this is a home network, I had to create this new by learning ISC DHCP, plus I took the opportunity to install ISC BIND as a DNS server – further learning required. Note, you don’t *need* these for iPXE to work, but it makes life easier in the long run. I may blog more about setting up iPXE at some point, but there are plenty of documented guides on how to do it.
And what is the point of this blog?
So, as I’ve already mentioned above, iPXE doesn’t have *that* great a set of documentation, particularly in regard to configuring images to boot. Most of these comes through gleaning facts from various sources, putting them together through trial, error and a little bit of prior experiencing and seeing if it works. Whilst I manage to get things to boot, I think its useful to share solutions with the wider community to make things easier to use. This is particularly true when trying to convert Linux live images into an iPXE boot set up, where you don’t necessarily want to or can’t load an entire ISO into memory. Hopefully others will find this a useful resource, and I’m going to try and use it as a library of knowledge on my own experiences of network booting.