DEV Community

mikeyGlitz
mikeyGlitz

Posted on • Edited on

The New Cluster

I had previously built a compute cluster using Raspberry Pi mini-computers. A drawback that I had encountered was the ARM architecture that the Raspberry Pi uses. A lot of application developers don't support ARMhf or ARMV7 architectures.

Finally got around to publishing what I've been working on with my #RaspberryPi #Docker #ansible #mylittlerackofpis https://t.co/IHXL4izGnn

— Michael Glitzos (@mikeyglitz) January 3, 2018

Recently I came across this post from Hackernoon which explained how to make a DIY cluster using affordable stick-pcs. I decided to try building a compute cluster for the second time, but using stick pcs which I was able to affordably find on the internet.

Mini Stick PC

My cluster will consist of the following nodes:

  • Intel BOXNUC5CPYH NUC - Master
  • 2x ACEPC W5 - Worker

The Intel NUC was a mini-computer which I had from a previous home lab set up. The NUC has the following specs:

  • Intel Celeron N3050 - 2 cores 2.16 Ghz
  • 8 GB DDR3
  • 128 GB SSD

Cluster

I've attached an additional 4TB USB HDD to for storage. I intend to use the master node as a network-attached storage (NAS) device as well.

The compute nodes have the following specs:

  • Intel Atom z8350 - 4 cores 1.8 GHZ
  • 2 GB DDR3
  • 32 GB Internal SD Card

Since the resources on these mini computers is limited, I'll be using k3s as a lightweight alternative to kubernetes.

Cluster Layout

I also found that I receive better performance and stability using Debian over Ubuntu.

Top comments (0)