Non-persisting worker-nodes

We want to create non-persisting worker-nodes without local storage

How can we install rancherOS, such that it can be booted via network and automatically connects to a rancher2.x-cluster?


What I have in mind is something like this:

  1. boot from official rancherOS.iso in some VM
  2. copy cloudconfig.yaml to VM (cluster-keys are hard-coded in cloud-config)
  3. ros install
  4. shutdown
  5. convert VM to bootable ISO
  6. register that ISO in iPXE
  7. boot via iPXE (repeat on all worker)

Can this work in theory?

Is there a better way to install my cloudconfig in a rancheros.iso?

When I tried it, I couldn’t convert my vdi to a bootable iso. I think it would be possible, but I don’t want to sink any more time into it, if it’s not possible to begin with.


Background:

We have a students-pool with ~15 machines (i7, 16GB RAM, no hdd). The students often work with confidential data and all use different OSes, so they each have their own hard-discs. All machines are equipped with slots for removable hard disks.

Most of the time (especially at night/during weekends) those machines are off. Also we have scientific workloads, that require lots of processing for a couple of days. So it would be neat if we could utilize those idle machines easily.