Files
home-network-k3s-impl/apps/vms/unifi/20-datavolume.yaml
T
jakman85andClaude Opus 4.8 df5979fceb fix(k3s): enable nonroot-devices so CDI can import to Block-mode PVCs
The UniFi VM boot-disk DataVolume (RWX + Block on longhorn-vm) failed to import:
CDI's non-root importer (UID 107, cap-drop ALL) hit
`blockdev: cannot open /dev/cdi-block-volume: Permission denied`.

Kubernetes only applies fsGroup ownership to *filesystem* volumes — it never
chowns a raw block device — and Longhorn's CSIDriver (fsGroupPolicy:
ReadWriteOnceWithFSType) skips fsGroup on RWX/Block volumes regardless, so the
device stays root:root. The CDI-documented fix is the container-runtime setting
`device_ownership_from_security_context`, exposed on k3s (>= v1.31.3+k3s1, this
cluster is v1.31.5) as `nonroot-devices: true`. It makes containerd set the
in-container device node's owner from the pod's securityContext.

Added to the Ansible-managed /etc/rancher/k3s/config.yaml alongside tls-san;
applied via the existing rolling `Restart k3s` handler (serial:1, kube-vip keeps
the API up, etcd holds 2/3). Node-scoped, security-neutral (pods stay non-root),
and preserves the longhorn-vm RWX/Block/migratable contract unchanged.

Co-Authored-By: Claude Opus 4.8 <[email protected]>
2026-08-26 15:24:07 -05:00

50 lines
2.5 KiB
YAML

---
# STAGING / DRAFT — see namespace.yaml header.
#
# CDI DataVolume — the UniFi VM boot disk. This is the FIRST real CDI *http* import on
# this cluster (the vm-demo/conformance path imports a containerdisk from a *registry*;
# this pulls the official Ubuntu 24.04 LTS server cloud image (qcow2) over https and CDI
# converts it onto the PVC).
#
# STANDALONE DataVolume (NOT a dataVolumeTemplate) so the disk OUTLIVES `kubectl delete vm`
# and survives destroy/rebuild (R-STATE-01) — same durability contract as vm-demo.
#
# STORAGE = longhorn-vm (the RWX/migratable StorageClass the storage workstream is adding):
# - accessModes: ReadWriteMany + volumeMode: Block => REQUIRED for KubeVirt LIVE MIGRATION
# (evictionStrategy: LiveMigrate in the VM). A ReadWriteOnce disk cannot live-migrate.
# - 60Gi: the Ubuntu cloud image virtual size is ~3.5Gi; cloud-init growpart/growroot
# expands the root fs to fill the disk on first boot. 60Gi leaves ample room for the
# UniFi datastore + OS + podman images + updates.
#
# GOTCHA / FLAG: a Block RWX DataVolume import needs CDI scratch space and the importer pod
# needs egress to cloud-images.ubuntu.com (see the CDI-import allow in 40-networkpolicy.yaml —
# the importer runs in THIS default-deny namespace, unlike the conformance test whose ephemeral
# ns has no baseline). No checksum is verified on an http source; pin a dated build for full
# reproducibility if drift matters (see README "Open questions").
#
# NODE PREREQ (Block-mode CDI): the non-root importer opens the raw block device, which k8s
# never chowns for it. The k3s nodes set `nonroot-devices: true`
# (infra/ansible/roles/k3s_server) so containerd hands the device to the pod's UID/GID —
# without it the import dies with `blockdev: cannot open /dev/cdi-block-volume: Permission denied`.
apiVersion: cdi.kubevirt.io/v1beta1
kind: DataVolume
metadata:
name: unifi-rootdisk
namespace: unifi
labels:
app.kubernetes.io/name: unifi
spec:
source:
http:
# Ubuntu 24.04 LTS (noble) server cloud image, amd64, qcow2. "releases/.../release/"
# tracks the latest 24.04 point release. Pin to a dated build dir for reproducibility.
url: "https://cloud-images.ubuntu.com/releases/noble/release/ubuntu-24.04-server-cloudimg-amd64.img"
storage:
accessModes:
- ReadWriteMany # RWX => migratable (paired with evictionStrategy: LiveMigrate)
volumeMode: Block # block mode is the migratable/perf default for VM disks
storageClassName: longhorn-vm
resources:
requests:
storage: 60Gi