Skip to content

Install Fabric

Prerequisites

  • A machine with access to the Internet to use Fabricator and build installer with at least 8 GB RAM and 25 GB of disk space
  • An 16 GB USB flash drive, if you are not using virtual media
  • Have a machine to function as the Fabric Control Node. System Requirements as well as IPMI access to it to install the OS.
  • A management switch with at least 1 10GbE port is recommended
  • Enough Supported Switches for your Fabric

Overview of Install Process

This section is dedicated to the Hedgehog Fabric installation on bare-metal control node(s) and switches, their preparation and configuration. To install the VLAB see VLAB Overview.

Download and install hhfab following instructions from the Download section.

The main steps to install Fabric are:

  1. Install hhfab on the machines with access to the Internet
    1. Prepare Wiring Diagram
    2. Select Fabric Configuration
    3. Build Control Node configuration and installer
  2. Install Control Node
    1. Insert USB with control-os image into Fabric Control Node
    2. Boot the node off the USB to initiate the installation
  3. Prepare Management Network
    1. Connect management switch to Fabric control node
    2. Connect 1GbE Management port of switches to management switch
  4. Prepare supported switches
    1. Ensure switch serial numbers and / or first management interface MAC addresses are recorded in wiring diagram
    2. Boot them into ONIE Install Mode to have them automatically provisioned

Build Control Node configuration and Installer

Hedgehog has created a command line utility, called hhfab, that helps generate the wiring diagram and fabric configuration, validate the supplied configurations, and generate an installation image (.img or .iso) suitable for writing to a USB flash drive or mounting via IPMI virtual media. The first hhfab command to run is hhfab init. This will generate the main configuration file, fab.yaml. fab.yaml is responsible for almost every configuration of the fabric with the exception of the wiring. Each command and subcommand have usage messages, simply supply the -h flag to your command or sub command to see the available options. For example hhfab vlab -h and hhfab vlab gen -h.

HHFAB commands to make a bootable image

  1. hhfab init --wiring wiring-lab.yaml
  2. The init command generates a fab.yaml file, edit the fab.yaml file for your needs
    1. ensure the correct boot disk (e.g. /dev/sda) and control node NIC names are supplied
  3. hhfab validate
  4. hhfab build --mode iso
    1. An ISO is best suited to use with IPMI based virtual media. If desired an IMG file suitable for writing to a USB drive, can be created by passing the --mode usb option. ISO is the default.

The installer for the fabric is generated in $CWD/result/. This installation image is named control-1-install-usb.iso and is 7.5 GB in size. Once the image is created, you can write it to a USB drive, or mount it via virtual media.

Write USB Image to Disk

This will erase data on the USB disk.

  1. Insert the USB to your machine
  2. Identify the path to your USB stick, for example: /dev/sdc
  3. Issue the command to write the image to the USB drive
    • sudo dd if=control-1-install-usb.img of=/dev/sdc bs=4k status=progress
  1. Plug the drive into the computer
  2. Open the terminal
  3. Identify the drive using diskutil list
  4. Unmount the disk diskutil unmount disk5, the disk is specific to your environment
  5. Write the image to the disk: sudo dd if=./control-1-install-usb.img of=/dev/disk5 bs=4k status=progress

There are utilities that assist this process such as etcher.

Install Control Node

This control node should be given a static IP address. Either a lease or statically assigned.

  1. Configure the server to use UEFI boot without secure boot

  2. Attach the image to the server either by inserting via USB, or attaching via virtual media

  3. Select boot off of the attached media, the installation process is automated

  4. Once the control node has booted, it logs in automatically and begins the installation process

    1. Optionally use journalctl -f -u flatcar-install.service to monitor progress

  5. Once the installation is complete, the system automatically reboots.

  6. After the system has shutdown but before the boot up process reaches the operating system, remove the USB image from the system. Removal during the UEFI boot screen is acceptable.

  7. Upon booting into the freshly installed system, the fabric installation will automatically begin

    1. If the insecure --dev flag was passed to hhfab init the password for the core user is HHFab.Admin!, the switches have two users created admin and op. admin has administrator privileges and password HHFab.Admin!, whereas the op user is a read-only, non-sudo user with password HHFab.Op!.
    2. Optionally this can be monitored with journalctl -f -u fabric-install.service

  8. The install is complete when the log emits "Control Node installation complete". Additionally, the systemctl status will show inactive (dead) indicating that the executable has finished.

Configure Management Network

The control node is dual-homed: it connects to two different networks, which are called management and external, respectively, in the fab.yaml file. The management network is for controlling the switches that comprise the fabric. It can be a simple broadcast domain with layer 2 connectivity. The management network is not accessible to machines or devices not associated with the fabric; it is a private, exclusive network. The control node connects to the management network via a 10 GbE interface. It runs a DHCP server, as well as a small HTTP server.

The external network allows the user to access the control node via their local IT network. It provides SSH access to the host operating system on the control node.

Fabric Manages Switches

Now that the install has finished, you can start interacting with the Fabric using kubectl, kubectl fabric and k9s, all pre-installed as part of the Control Node installer.

At this stage, the fabric hands out DHCP addresses to the switches via the management network. Optionally, you can monitor this process by going through the following steps: - enter k9s at the command prompt - use the arrow keys to select the pod named fabric-boot - the logs of the pod will be displayed showing the DHCP lease process - use the switches screen of k9s to see the heartbeat column to verify the connection between switch and controller. - to see the switches type :switches (like a vim command) into k9s