Skip to content

Reference

What belongs here

Reference is a map, consulted mid-task and never read through. It describes the machinery and does nothing else: no instruction, no persuasion, no worked examples.

Admission test:

  • The reader arrives already knowing what they are looking for.
  • It is structured to match the thing it describes — one section per storage class, one row per policy — not structured as an argument.
  • It is neutral. The moment a page starts justifying a choice, that part belongs in Explanation.

Prefer generated reference

Reference rots faster than anything else here and benefits least from prose. The app inventory, ingress hosts, namespaces and storage classes are all derivable from the manifests — generate them in CI rather than typing them.

Available now

  • Applications — every app, its namespace, its Flux Kustomization and its hostnames (generated)
  • Admission policies — every Kyverno policy wherever it lives, what it does, and which rules only warn (generated)
  • Flux Kustomizations — every Kustomization by layer, with interval, prune, wait and dependsOn (generated)
  • Helm releases — every release, its chart and source, and where its values come from (generated)
  • Ingress and certificates — the three classes, the two issuers, and what external-dns publishes
  • Storage classes — capabilities, access modes, node affinity and measured ceilings
  • Annotations and labels — keys this cluster gives meaning to, and two that look load-bearing but are not
  • Charts and images — the two sibling repositories, what they hold, and where their generated reference lives
  • Hardware — the nodes, their disks, and which physical device backs each volume group
  • UPS Modbus register map — every register confirmed against the front panel, and how to switch Modbus on