Files
home-network-k3s-impl/platform/kubevirt/cdi-cr.yaml
jakman85andClaude Opus 4.8 7a3da96d9f fix(cdi): single-replica scratch class + 40Gi UniFi disk (fix insufficient-storage import)
The UniFi Block import got past the device-permission fix but then failed
ReplicaSchedulingFailure / "insufficient storage": CDI sizes scratch ~= the disk
virtual size and, on the default 3-replica `longhorn` class, the ~60Gi scratch x3
(189G) on top of the disk's x3 (180G) overflowed the Longhorn tier (~125G/node).

- platform/longhorn/storageclasses.yaml: NEW `longhorn-scratch` StorageClass
  (numberOfReplicas: 1) — CDI scratch is ephemeral/throwaway, replicating it 3x is
  waste; if a node dies mid-import CDI just retries.
- platform/kubevirt/cdi-cr.yaml: spec.config.scratchSpaceStorageClass ->
  longhorn-scratch.
- apps/vms/unifi/20-datavolume.yaml: boot disk 60Gi -> 40Gi (UOS min is 25Gi;
  40Gi is comfortable for a home site and lighter on the replicated tier).

Co-Authored-By: Claude Opus 4.8 <[email protected]>
2026-08-26 15:47:02 -05:00

29 lines
1.4 KiB
YAML

---
# CDI CR — Containerized Data Importer, provisions/imports VM disks onto Longhorn PVCs (ADR-0005).
# CONSUMES the cdi.kubevirt.io CRD registered by the CDI operator, so — like the KubeVirt CR — it is
# applied by the `platform-config` Flux Kustomization (dependsOn: platform), not in the controllers
# pass (F-MON1). VM disks then get the same 3-replica Longhorn durability + backup story as container
# volumes (docs/04).
apiVersion: cdi.kubevirt.io/v1beta1
kind: CDI
metadata:
name: cdi
namespace: cdi
annotations:
homelab.keithjackson.dev/adr: ADR-0005
spec:
# ImagePullPolicy for CDI's own importer/uploader images.
imagePullPolicy: IfNotPresent
infra: {}
workload: {}
config:
# CDI's transient scratch space (used while converting/decompressing an imported image) is carved
# from a dedicated SINGLE-replica Longhorn class — scratch is ephemeral/throwaway, so replicating
# it 3x is waste and, for a large VM disk, the 3x scratch on top of the 3x final disk exhausts the
# Longhorn tier (ReplicaSchedulingFailure). See platform/longhorn/storageclasses.yaml. (docs/04, ADR-0003.)
scratchSpaceStorageClass: longhorn-scratch
# Reserve headroom on the filesystem-mode PVC for qcow2 metadata so an import never wedges a volume
# at 100%. 0.055 is CDI's documented default; pinned here so it is explicit and reviewable.
filesystemOverhead:
global: "0.055"