The Kubernetes side of the homelab runs on Talos and has gradually become one of the central pieces of the wider environment.

The cluster is managed with Flux GitOps, SOPS and age for secrets, Cilium as the CNI, MetalLB for service exposure, and cert-manager for certificate handling. That gives a clean operational model, but it also creates a lot of interesting edge cases around upgrades, node dependencies, and hardware quirks.

Recent work has included:

  • assessing the Talos 1.12.0 upgrade path
  • working through Intel GPU CDI configuration for Plex transcoding
  • handling wake-on-LAN limitations on X520-based interfaces
  • keeping cluster shutdown behaviour aligned with the UPS automation

This section should eventually grow into a set of focused notes rather than one giant overview, because the value is usually in the awkward details.