Skip to content

README.md

Latest commit

 

History

History
42 lines (28 loc) · 2.93 KB

README.md

File metadata and controls

42 lines (28 loc) · 2.93 KB

raspberry-pi-netboot - Boot your Pi off iSCSI

raspberry-pi-netboot contains a Packer template for building diskless Raspberry Pi images and a build tool for provisioning them. It's meant to be run on an ARM machine, because this is a personal project and I happen to have one. This repository also has no license because I used a lot of blog posts to put it together, so do what you want with it.

Special thanks to the following blog posts for inspiration:

Getting started

To start, build a golden image for the nodes by running make image. This will write a Raspberry Pi OS image to images/raspios.img. If you don't have an ARM machine to build on and want to make use of binfmt_misc functionality in the Packer builder, edit the Makefile to remove the DONT_SETUP_QEMU environment variable declaration.

Once you have an image, define your build configuration in a JSON file in configs/workspace.json. This file defines fleet-level settings like OS image and network storage locations. Then, define your instance configurations as JSON files in the configs/instances directory. These files define instance-level settings like instance ID, iSCSI IQNs, MAC address, and authentication information. For more information about these values, see provision/src/config.rs.

To provision the nodes, run make provision. This will perform the following actions for each node:

  • Write bootloader files to the node's /boot/firmware mount point
  • Create a filesystem at the node's iSCSI target and write root filesystem files to the node's / mount point
  • Configure /etc/fstab and the kernel command line to boot via iSCSI
  • Configure SSH to disallow passwordless auth and allow root login with the provided public key

The exact build steps are located in provision/src/steps.rs. The graph defining build steps is located in provision/src/lib.rs.

Notes

This provisioning flow makes some assumptions about the deployment environment. In particular, it assumes:

  • Each machine is a Raspberry Pi 4
  • Each machine is configured as a network boot client
  • A DHCP server manages host configuration
  • The DHCP server advertises the location of an NTP server using DHCP option 42
  • The DHCP server advertises the location of the TFTP server using DHCP option 66
  • The TFTP server is running on the same subnet as the Raspberry Pi