Rabbit Launch: Loading up the NUC cluster with a usable operating system

As you saw in my 3D Printing series, after years of pondering a 3D printer, I was finally inspired to buy one when a pile of clusters came up on eBay from the defunct rabb.it video streaming service. In this series, I’ll take you through turning a rabbit door into some useful computing resources.

You can do something similar even after the clusters are sold out; a lot of people have probably bought the clusters and ended up not using them, so you’ll see boards on eBay or local marketplaces… or you can adjust the plans here to other models.

The previous installment, Rabbit Reorganization, can be found here. Stay tuned for more coverage.

Update December 2021: This post languished in the drafts folder for about a year. I’ve updated links, and I’ll be reporting on some changes since the October 2020 launch of this cluster soon.

Let’s NUC this cluster out

  1. Install memory, SATA cable, and SSD
  2. Upgrade BIOS and set some annoying settings
  3. Install your operating system
  4. Set up central control

Install memory, SATA cable, and SSD

This is the least interesting part of the process, but you’ll need to do it before you can install an OS.

Start by loading the SODIMM of your choice onto the board. If you’re using an SSD like I am, you’ll connect the SATA cable to the black SATA connector next to the front USB stack, and the power cable to the beige connector perpendicular to the SATA connector. If you’re using a standard SD card, plug it into the SD slot to the right (as shown with the MAC address label). If you’re going with netboot (local storage? where we’re going we don’t need local storage!), just connect your network cable.

Upgrade BIOS and set some annoying settings

I created a bootable FreeDOS USB drive with Rufus, a common free software product used to create bootable USB media from ISOs (think Linux, Windows, etc). From there, we get the latest BIOS from Intel’s Download Center and place the file on the bootable drive. (Further BIOS instructions available on Intel Support.)

As of December 2021, 0079 is the latest BIOS, released April 20, 2020. You’ll need to search for NUC5PPYH even though the board’s model is PPYB.

Connect a monitor and keyboard, plug in the bootable drive, and apply power (or just reset the board). Use the F7 key to go into onboard flash update and load the BIOS file from the flash drive, or choose to boot from the flash drive and use the DOS-style flasher from there.

When you’ve done the upgrade and the system has rebooted, go into the BIOS with the F2 key and choose BIOS default values. Then go into the menus to enable all the USB ports (for some reason the default is to enable ports 1-3, leaving physical port 4 and header ports 5-6 disabled) as well as the SATA port if you are using that for storage. I’d also check the boot order (move net boot down in preference or disable outright if you don’t plan to use it). You can choose other settings as desired, and then press F10 to save and reboot.

Install your operating system

The easiest way to roll out the NUC side of the door would be to netboot an installation infrastructure like Cobbler. One of the first things I did when I went to work for the Mouse 10 years ago was setting up Cobbler for a deployment of RHEL 5.5.

Sure enough, Cobbler is still a thing, with very recent updates. I was able to get partway there this time and then, after several dozen runs to the garage and back to power cycle nodes, I gave up and installed from local media.

For CentOS 8, I did a manual install booting from a Rufus-created USB drive, with the SSD installed. I configured my storage and network options by hand, as well as user and root credentials. This left an “anaconda.ks” kickstart file on the installed system, which I copied to a second flash drive.

For the additional systems, I plugged both the CentOS 8 installer and the drive with the kickstart file into the NUCs and booted from USB. I ran into some strange storage issues with the drive not being blank, despite having chosen the kickstart option. Ideally, you would boot from the USB installer, it would find your kickstart config, and just roll out the software without intervention from there.

After that, if your DHCP server doesn’t assign hostnames you like, you can go in and set hostnames with hostnamectl or the like.

Set up central control

If you use a configuration management platform like Ansible, Puppet, Chef, cfengine, or the like, you’ll want to set those up at this point.

I’ve gone with the lightweight method so far, with shared SSH keys from a management host (an Intel NUC with CentOS on it, originally intended to be the cobbler server).

Use ssh-keygen to create your key files, and then ssh-copy-id can be used to push out the keys to your hosts. Then look into a more manageable option.

Where do we go from here?

As I finish this post in December 2021, a year after the original build, I’m looking at going back and making a few changes to the cluster to bring it up as a Kubernetes platform.

With the demise of CentOS as many of us know it, I’m planning to replace the installed OSes with Ubuntu LTS. I’m planning to test out some cryptocurrency cpu-based mining, and run Kubernetes platform(s) on it as well, and bring my second door up to speed (the RAM has been sitting in a box in the living room for a year now).

There’s a chance I’ll even do some lightweight Chia farming, using either bus-powered USB hard drives or some of the extra power connectors from the fused expanders for standard Seagate externals.

For those of you who have bought and built up these doors, what did you do with them? Feel free to share details and blog post links in the comments. I’ll put interesting ones into the body of this post as I see them.

7 thoughts on “Rabbit Launch: Loading up the NUC cluster with a usable operating system

  1. Pingback: Rabbit Reorganization: Building low power clusters from a rabb.it door | rsts11 – Robert Novak on system administration

  2. I’ve been modding my rabbit cluster (just NUCs, not nvidia jetsons) to fit into an old ITX case. It’s a tight fit, and especially tricky with all of the SATA hard drives.

    Liked by 1 person

    • If you have photos or a post about this effort, I’d be interested. I was pondering 3d printed 5.25″ drive tray mounts (and I have an Antec Twelve Hundred that would be great for this) but never got around to trying it.

      Like

      • That looks pretty cool. Having the drives outside the stack is a good idea. I may go that way on the next version… I have some 4x 2.5″ pluggable bays that I could use to front-end them in a case or just bolt onto the stack.

        Like

      • I figured through the iteration process of bootstrapping the cluster, if a proper PXE setup failed me, I could always just eject the drives and image them outside the system – in particular without having to shut the whole cluster down (just the individual node), and without ever having to go in with a keyboard/display/usb boot drive.

        That said – getting to that ideal setup in such a small space has been challenging, but I do like the form factor. Here’s another similar setup: https://imgur.com/a/IedEB7J

        Like

  3. Pingback: Pi in the sky: Seven tips for finding the single board computer of your dreams | rsts11 – Robert Novak on system administration

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.