0

I'm trying to learn the basics of Linux clustering so I started designing a really humble cluster:

  • 6 worker nodes (Libre Computer La Frite | Cortex-A53 @ 1.2 GHz | 1GB RAM)
  • 1 master node (Raspberry Pi 4 Model B | Cortex-A72 @ 1.5 GHz | 2GB RAM)
  • 16-port Gbit Ethernet switch
  • 500GB SSD shared through NFS on the network

I'm planning to only run k3s on it and because of the diskless nature of the workers SBC boards, I was wondering which kind of "root filesystem structure" i should adopt for the cluster, on the NFS storage.

For e.g. :

  • single-system image with the same shared root fs across all nodes

  • separate root fs for each node (/opt/node1/* , /opt/node2/* , ...)

    and so on...

Thanks to everyone who will reply!

phreq
  • 1
  • 2
  • shared root is certainly easier to keep all machines upgraded/configured in sync and saves on disk space requirements BUT a) there will be many complications if the rootfs needs to be writable by each node (so try to avoid that if possible); b) you'll need to write some custom scripts to ensure that each node has its own hostname and k3s node token (and probably other per-node variables) - these could be derived from the node's IP address (or MAC address) with a simple lookup table. – cas Apr 01 '23 at 01:09
  • BTW, there's no problem with your question being "too advanced for a beginner" - good on you for trying and being willing to learn. The problem is that your question is too large in scope for a question on this site. U&L isn't really suited to questions like that. – cas Apr 01 '23 at 01:12
  • Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. – Community Apr 01 '23 at 14:17
  • Thank you very much @cas, do you know a place that could suit better this question? – phreq Apr 01 '23 at 15:31
  • sorry, no. SE (and similar question & answer) sites are mostly oriented towards answering fairly narrowly focused questions - the details, rather than big picture design. If you can rework your question into something more focused it could work here. Otherwise, maybe look for tutorials or blog posts or youtube videos etc on k3s (btw, i don't know k3s at all, but i have built diskless workstations on NFS, and both PBS & slurm-based HPC clusters with shared data storage in the past, as well as openstack clusters). – cas Apr 02 '23 at 00:07
  • "learn the basics...diskless nodes....run k3s...500GB SSD shared" - perhaps you should try something less ambitious to start with. That's at least 3 very advanced problems to solve at the same time. – symcbean Apr 02 '23 at 01:12

0 Answers0