KNEL/k8s — K8S platform body of work

Kubernetes platform for the KNEL fleet: k3s cluster lifecycle (install/join/ verify/wipe), platform decisions (vcluster vs Rancher), OIDC/RBAC, storage, resource management, security levels, multi-tenant workloads.

Provenance: split out of KNEL/PFVCluster k8s/ on 2026-09-05 as part of the per-domain repo split (#769).

Systems of record

what	where
Redmine project	k8s-platform — master ticket #699
Active tickets	#700 Rancher+OIDC, #701 vcluster slicing, #702 gitops
Discourse	https://community.turnsys.com/t/336
Cluster	pfv-k8s (k3s), canonical architecture doc: docs/ARCHITECTURE.md

Layout

• install-cp.sh, join-servers.sh, join-workers.sh, post-setup.sh — k3s lifecycle
• verify.sh, probe-nodes.sh, wipe.sh — health + teardown
• proxmox-tuning.sh — VM tuning for k8s nodes
• code/ — workload manifests (nvidia device plugin)
• docs/ — ARCHITECTURE, DISTRO-DECISION
• kubeconfig.example.yaml — shape only. NEVER commit a real kubeconfig (.gitignore enforces; real kubeconfigs live out of repo).

Status

See STATUS.md. Platform direction (vcluster vs Rancher vs both) is being deliberated on Discourse — decision lands as an ADR + [#699].